Searching for a Listed Infrastructure Asset Class

A Publication of the EDHEC Infrastructure Institute-Singapore

Mean-variance spanning tests of 22 listed infrastructure proxies

June 2016

infra Singapore Infrastructure Investment Institute Table of Contents

Executive Summary ...... 4

1 Introduction ...... 11

2 Literature Review ...... 14

3 Approach ...... 18

4 Methodology ...... 21

5 Data ...... 24

6 Results ...... 34

7 Conclusions ...... 50

References ...... 61

About the EDHEC Infrastructure Institute-Singapore ...... 63

Infrastructure Research Publications at EDHEC ...... 68

The author would like to thank Noel Amenc, Lionel Martellini, Timo Välilä and two anonymous reviewers for useful comments and suggestions. This study presents the author’s views and conclusions which are not necessarily those of EDHEC Business School. Searching for a listed infrastructure asset class - June 2016

About the Authors

Frédéric Blanc-Brude is Director of the EDHEC Infrastructure Institute–Singapore and represents EDHEC Business School on the Advisory Board of the Global Infrastructure Facility of the World Bank. He holds a PhD in Finance (King’s College London), and degrees from London School of Economics, the Sorbonne and Sciences Po Paris.

Tim Whittaker is an Associate Research Director at EDHEC Infras- tructure Institute-Singapore and Head of Data Collection. He holds a Master of Business (Financial Management) an, Bachelors of Economics and Commerce from the University of Queensland, and a PhD in Finance from Griffith University.

Simon Wilde is a Ph.D. candidate at the University of Bath, UK and a Senior Managing Director at Macquarie Capital in London. He is also part-time Senior Lecturer at UWE Bristol Business School. He holds degrees from the London School of Economics and the Universities of Cambridge and Bristol.

A Publication of the EDHEC Infrastructure Institute-Singapore 3 Executive Summary

4 A Publication of the EDHEC Infrastructure Institute-Singapore Searching for a listed infrastructure asset class - June 2016

Executive Summary

In this paper, we ask the question: does conclusive evidence is not forthcoming focusing on listed infrastructure stocks in existing papers; create diversification benefits previously 2. Index providers have created dedicated unavailable to large investors already active indices focusing on this theme and a in public markets? number of active managers propose to invest in ”listed infrastructure” arguing This question arises from what we call that it does indeed constitute a unique the ”infrastructure investment narrative” asset class; (Blanc-Brude, 2013), a set of investment 3. Listed infrastructure stocks are often beliefs commonly held by investors about used by investors to proxy investments the investment characteristics of infras- in privately held (unlisted) infrastructure tructure assets. equity, but the adequacy of such proxies remains untested. In this narrative, the ”infrastructure asset class” is less exposed to the business cycle The existence of a distinctive listed infras- because of the low price-elasticity of infrastructure effect in investors’ portfolio would tructure services. Furthermore, the value of support these views. In the negative, if these investment is expected to be mostly this effect cannot be found, there is little determined by income streams extending to expect from listed infrastructure equity far into the future, and should thus be less from an asset allocation (risk/reward optimi- impacted by current events. sation) perspective and maybe even less to learn from public markets about the According to this intuition, listed infras- expected performance of unlisted infrastructure may provide diversification tructure investments. benefits to investors since they are expected to exhibit low return covariance with other financial assets. In other words, Testing 22 proxies of listed listed infrastructure is expected to exhibit infrastructure sufficiently unique characteristics to be We test the impact of adding 22 different considered an ”asset class” in its own right. proxies of ”listed infrastructure” to the portfolio of a well-diversified investor using Empirically, there are at least three reasons mean-variance spanning tests. We focus on why this view requires further examination: three definitions of ”listed infrastructure” as an asset selection scheme: 1. Most existing research on infrastructure has used public equity markets to infer 1. A ”naïve”, rule-based filtering of stocks findings for the whole infrastructure based on industrial sector classifications investment universe, but robust and and percentage income generated from

A Publication of the EDHEC Infrastructure Institute-Singapore 5 Searching for a listed infrastructure asset class- June 2016

Executive Summary

pre-defined infrastructure sectors (nine allocation defined either by traditional proxies); asset class or by factor exposure after the 2. Existing listed infrastructure indices GFC and only one is not already spanned designed and maintained by index both pre- and post-GFC; providers (twelve proxies); 4. Building baskets of stocks on the basis of 3. A basket of stocks offering a pure their SIC code and sector-derived income exposure to several hundred underlying fails to generate a convincing exposure projects that correspond to a well- to a new asset class. known form of infrastructure investment 5. Hence, benchmarking unlisted infras- defined – in contrast with the two tructure investments with thematic previous cases – in terms of long-term (industry-based) stock indices is unlikely public-private contracts, not industrial to be very helpful from a pure asset sectors (one proxy). allocation perspective i.e. the latter do not exhibit a risk/return trade-off or Employing the mean-variance spanning betas that large investors did not have tests originally described by Huberman and access to already. Kandel (1987) and Kan and Zhou (2012), we test the diversification benefits of these Overall, we do not find persistent evidence proxies of the listed infrastructure effect. to support the claims that listed infrastructure is an asset class. In other words, any ”listed infrastructure” effect was already There is no listed infrastructure spanned by a combination of capital market asset class instruments over the past 15 years in Global, Stylised findings include: US and UK markets.

1. Our 22 tests of listed infrastructure Defining infrastructure investments asa reveal little to no robust evidence of series of industrial sectors and/or tangible a ”listed infrastructure asset class” that assets is fundamentally misleading. We was not already spanned by a combi- find that such asset selection schemes do nation of capital market instruments and not create diversification benefits, whether alternatives, or by a factor-based asset reference portfolios are structured by tradi- allocation; tional asset classes or factor exposures. 2. The majority of test portfolios that improve the mean-variance efficient We conclude that what is typically referred frontier before the GFC fail to repeat this to as listed infrastructure, defined by SIC feat post-GFC. There is no evidence of code and industrial sector, is not an asset persistent diversification benefits; class or a unique combination of market 3. Of the 22 test portfolios used, only four factors, but instead cannot be persistently manage to improve on a typical asset distinguished from existing exposures in

6 A Publication of the EDHEC Infrastructure Institute-Singapore Searching for a listed infrastructure asset class - June 2016

Executive Summary

Figure 1: Illustration of the difference tests of mean-variance spanning of the FTSE Macquarie USA Infrastructure Index

(a) 2000-2014, asset class and factor-based reference Efficient Frontier January 2000 to December 2013 0.20 0.20 Efficient Frontier February 2007 to December 2013

0.15 0.15

0.10 0.10 High Yield ● ● ● US Equities ● FTSE Macquarie US Infra FTSE Macquarie US Infra

● Market

Returns ●

Real Estate Returns ● Term Corporate Bonds ● 0.05 ● Hedge Funds 0.05 ● Government Bonds ● World Ex−US Equities

● Default 0.00 0.00 ● Size

● Value

● Commodities −0.05 −0.05 0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.00 0.05 0.10 0.15 0.20 0.25 0.30 Standard Deviation Standard Deviation

Reference Assets Reference and Infrastructure Assets Reference Assets Reference and Infrastructure Assets

(b) 2000-2008, asset class and factor-based reference Efficient Frontier February 2007 to December 2008 0.25 0.20 Efficient Frontier January 2000 to December 2008

0.20

0.15 0.15

● Government Bonds 0.10 ● Default 0.10

● Corporate Bonds ● Value 0.05 ● FTSE Macq

0.00 0.05 ● Size ● Term

−0.05 ● Hedge Funds Returns Returns

−0.10 0.00 ● FTSE Macquarie US Infra ● Commodities ● High Yield −0.15 ● Europe Stocks −0.05 −0.20 ● US Equities ● World Ex−US Equities ● US Stocks

−0.25

−0.30

● Real Estate 0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 0.00 0.05 0.10 0.15 0.20 0.25 0.30 Standard Deviation Standard Deviation

Reference Assets Reference and Infrastructure Assets Reference Assets Reference and Infrastructure Assets

0.25

● US Stocks 0.15 0.20 ● Real Estate

● US Equities

● High Yield 0.15 ● FTSE Macquarie US Infra ● Europe Stocks 0.10 ● FTSE Macq

● World Ex−US Equities Returns 0.10 Returns

● Hedge Funds 0.05 ● Size ● Term

0.05 ● Corporate Bonds

● Government Bonds

0.00

● Value 0.00

● Commodities

−0.05 −0.05 0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.00 0.05 0.10 0.15 0.20 0.25 0.30 Standard Deviation Standard Deviation

Reference Assets Reference and Infrastructure Assets Reference Assets Reference and Infrastructure Assets

A Publication of the EDHEC Infrastructure Institute-Singapore 7 Searching for a listed infrastructure asset class- June 2016

Executive Summary

Figure 2: Mean-Variance Frontier of total returns PFI portfolio and reference portfolio, 2000- 2014

0.20 Efficient Frontier January 2000 to December 2013

0.15

● PFI Portfolio 0.10

● World Ex−UK Equities Returns ● UK Equities ● UK Gilts 0.05

● Hedge Funds

● Real Estate

0.00

● Commodities −0.05 0.00 0.05 0.10 0.15 0.20 0.25 0.30 Standard Deviation

Reference Assets Reference and Infrastructure Assets

investors’ portfolios, and that expecting the ”add-ons” to such approaches, making them emergence of a new or unique ”infras- completely ad hoc and unscientific. tructure asset class” by focusing on public equities selected on the basis of industrial sectors is unlikely to be very useful for Defining infrastructure differently investors. Our tests also tentatively suggest a more promising avenue to ”find infrastructure” in Figure 1 provides an illustration of these the public equity space: focusing on under- results in the case of the FTSE Macquarie lying contractual or governance structures Listed Infrastructure Index for the U.S. that tend to maximise dividend payout and market. pay dividends with great regularity, such as the public-private partnerships (PPPs) or Thus, asset owners and managers who use master limited partnerships (MLPs) models, the common ”listed infrastructure” proxies we find that the mean-variance frontier of to benchmark private infrastructure invest- a reference investor can be improved. ments are either misrepresenting (probably over-estimating) the beta of private infras- The answer to our initial question partly tructure, and usually have to include various depends on how ”infrastructure” is defined

8 A Publication of the EDHEC Infrastructure Institute-Singapore Searching for a listed infrastructure asset class - June 2016

Executive Summary

and understood as an asset selection ability of investors to enjoy these potential scheme. benefits unless the far larger unlisted infrastructure fund universe has similar charac- Under our third definition of infrastructure, teristics. which focuses on the relationship-specific and contractual nature of the infrastructure Future work by EDHECinfra aims to answer business, we find that listed infrastructure these questions in the years to come. may help identify exposures that have at least the potential to persistently improve portfolio diversification on a total return basis, as figure 6 illustrates. This effect is driven by the regularity and the size of dividend payouts compared to other corporations, infrastructure or not.

What determines this ability to deliver regular and high dividend payouts is the contractual and governance structure of the underlying businesses, not their belonging to a given industrial sector. Bundles of PPP project companies or MLPs behave differently than regular corporations i.e. their ability to retain and control the free cash flow of the firm is limited and they tend to make large equity payouts. In the case if PPP firms, as Blanc-Brude et al. (2016) show, they also pay dividends with much greater probability than other firms.

Going beyond sector exposures and focusing on the underlying business model of the firm is more likely to reveal a unique combination of underlying risk factors.

However, it must be noted that the relatively low aggregate market capitalisation of listed entities offering a ”clean” exposure to infrastructure ”business models” as opposed to ”infrastructure corporates” may limit the

A Publication of the EDHEC Infrastructure Institute-Singapore 9 1. Introduction

10 A Publication of the EDHEC Infrastructure Institute-Singapore Searching for a listed infrastructure asset class - June 2016

1. Introduction

In this paper, we ask the question: does conclusive evidence is not forthcoming focusing on listed infrastructure stocks in existing papers; create diversification benefits previously 2. Index providers have created dedicated unavailable to large investors already active indices focusing on this theme and a in public markets? number of active managers propose to invest in ”listed infrastructure” arguing This question arises from what we call that it does indeed constitute an asset the ”infrastructure investment narrative” class in its own right, worthy of an (Blanc-Brude, 2013), a set of investment individual allocation; beliefs commonly held by investors about 3. Listed infrastructure stocks are often the investment characteristics of infras- used by investors to proxy investments tructure assets. in privately-held (unlisted) infrastructure but the adequacy of such proxies remains In this narrative, the ”infrastructure asset untested. class” is less exposed to the business cycle because of the low price-elasticity of infras- The existence of a distinctive listed infrastructure services. Furthermore, the value of tructure effect in investors’ portfolio would these investment is expected to be mostly support these views. In the negative, if determined by income streams extending this effect cannot be found, there is little far into the future, and should thus be less to expect from listed infrastructure equity impacted by current events. from an asset allocation (risk/reward optimi- sation) perspective and maybe even less According to this intuition, infrastructure to learn from public markets about the investments may provide diversification expected performance of unlisted infras- benefits to investors since they are expected tructure investments. to exhibit low return covariance with other financial assets, as well as a degree of We test the impact of adding 22 different downside protection. In other words, infras- proxies of public infrastructure stocks to the tructure investments are expected to exhibit portfolio of a well-diversified investor using sufficiently unique characteristics to be mean-variance spanning tests. We focus on considered an ”asset class” in its own right. three definitions of ”listed infrastructure” as an asset selection scheme: Empirically, there are at least three reasons why this view requires further examination: 1. A ”naïve”, rule-based filtering of stocks based on industrial sector classifications 1. Most existing research on infrastructure and percentage income generated from has used public equity markets to infer pre-defined infrastructure sectors (nine findings for the whole infrastructure proxies); investment universe, but robust and

A Publication of the EDHEC Infrastructure Institute-Singapore 11 Searching for a listed infrastructure asset class- June 2016

1. Introduction

2. Existing listed infrastructure indices tently improve portfolio diversification. In designed and maintained by index other words, going beyond sector exposures providers (twelve proxies); and focusing on the underlying business 3. A basket of stocks offering a pure model of the firm is more likely to reveal exposure to several hundred underlying a unique combination of underlying risk projects that correspond to a well- factors. known form of infrastructure investment defined – in contrast with the two The rest of this paper is structured as previous cases – in terms of long-term follows: section 2 briefly reviews existing public-private contracts, not industrial research on the performance of listed sectors (one proxy). infrastructure. Section 3 details our approach, while sections 4 and 5 present Overall, we do not find persistent evidence our choice of methodology and data, to support the claims that listed infras- respectively. The results of the analysis are tructure provides diversification benefits. reported in section 6. Finally, Section 7 In other words, any ”listed infrastructure” discusses our findings and their impli- effect was already spanned by a combi- cations to better define and benchmark nation of capital market instruments over infrastructure equity investments. the past 15 years in Global, US and UK markets.

We show that listed infrastructure, as it is traditionally defined (by their SIC code and industrial sector), is not an asset class or a unique combination of market factors, but instead cannot be distinguished from existing exposures in investors’ portfolios.

We also find that the answer to our question partly depends on how ”infrastructure” is defined and understood as an asset selection scheme.

Hence, under our third definition of infrastructure, which focuses on the relationship- specific and contractual nature of the infrastructure business, we find that listed infrastructure may help identify exposures that have at least the potential to persis-

12 A Publication of the EDHEC Infrastructure Institute-Singapore 2. Literature Review

A Publication of the EDHEC Infrastructure Institute-Singapore 13 Searching for a listed infrastructure asset class- June 2016

2. Literature Review

In Fabozzi and Markowitz (2011, p. 16), 2.1 Rule-based listed asset classes are defined as homogenous infrastructure portfolios investments with comparable character- The first group of papers simply examines istics, driven by similar factors, including a those stocks that are classified under a set common legal or regulatory structure, thus list of infrastructure activities and derive correlating highly with each other. a certain proportion of their income from these activities (see Newell and Peng (2007), As a direct result of this definition, the Finkenzeller et al. (2010), Newell and Peng combination of two or more asset classes (2008), Newell et al. (2009), Rothballer and can be expected to create diversifi- Kaserer (2012a) and Bitsch (2012)). cation benefits due to the limited return covariance of each group of assets. The findings from these studies suggest considerable heterogeneity in ”listed infras- The question of whether listed infrastructure”. Newell and Peng (2007) report tructure is an asset class and a good proxy that listed Australian infrastructure exhibits of a broader universe of privately-held higher returns, but also higher volatility infrastructure equity has been discussed in than equity markets. They find higher previous research. The approach taken in the Sharpe ratios than the market and low but literature has been to define infrastructure growing correlations over time with market in terms of industrial categories: roads and returns. Finkenzeller et al. (2010) find similar airports can seem rather alike as businesses results. when compared with automotive factories or financial services. Hence, following the The work of Newell and Peng (2008) definition given above, they can expected finds that in the U.S., infrastructure to form a relatively homogeneous group of (ex-utilities) under-performs stocks and stocks – a potential asset class – compared bonds over the period from 2000 to to other segments of the economy. 2006, while utilities outperform the market. Rothballer and Kaserer (2012a) find that Existing studies can be organised in two infrastructure stocks exhibit lower market groups. First, papers applying rule-based risk than equities in general but not stock selection schemes focusing on what is lower total risk i.e. they find high idiosyn- traditionally understood as ”infrastructure” cratic volatility. i.e. a collection of industrial sectors. Second, papers that employ listed infrastructure They also report significant heterogeneity in indices created by a number of index the risk profiles of different infrastructure providers. sectors with an average beta of 0.68 but with variation between sectors. For the utility, transport and telecom companies, the average betas were 0.50, 0.73 and 1.09,

14 A Publication of the EDHEC Infrastructure Institute-Singapore Searching for a listed infrastructure asset class - June 2016

2. Literature Review

1 1 - Using the same sample than respectively. Bitsch (2012) finds that infras- what qualifies as ”infrastructure” and apply Rothballer and Kaserer (2012a), Rödel tructure vehicles are priced using a high risk a market-capitalisation weighing scheme. and Rothballer (2012), examine the inflation hedging ability of infras- premium in part because of – he argue – They are ad hoc as opposed to rule-based tructure. They find no evidence to suggest complex and opaque financial structuring, because index providers pick and choose infrastructure exhibits a greater ability to hedge inflation risks than information asymmetries with managers, which stocks should be included in each listed equities. Even restricting their regulatory and political risks. infrastructure index. sample to firms with assumed strong monopoly characteristics does not yield a statistically significant result. These findings are in line with the results Using such indices, Bird et al. (2014) and of several industry studies suggesting that Bianchi et al. (2014) find that infrastructure the volatility of infrastructure indices is on exhibits similar returns, correlations and par with equities and real estate, but that tail-risks than the market, with a marginally market correlation is relatively low (Colonial higher Sharpe ratio, driven by what could be First State Asset Management, 2009; RREEF described as a ’utility tilt’. 2007). Other recent studies on the performance of The conclusions from this strand of liter- infrastructure indices by Peng and Newell ature are limited. ”Infrastructure” stocks are (2007), Finkenzeller et al. (2010), Dechant founds to have higher Sharpe ratios in and Finkenzeller (2013) and Oyedele et al. some cases but the statistical significance (2014) also report potential diversification of this effect is never tested. Overall, benefits, but none examine whether these rule-based infrastructure stocks selection are statistically or economically significant. schemes lead to either anecdotal (small For example, Peng and Newell (2007) and sample) or heterogenous results, which do Oyedele et al. (2014) compare Sharpe ratios not support the notion of an independent but provide no statistical tests to support asset class. their conclusions.

Idzorek and Armstrong (2009) provide the 2.2 Ad hoc listed infrastructure only study of the role of listed infrastructure indices in a portfolio context. The authors create an A second group of studies uses infras- infrastructure index by combining existing tructure indices created by index providers industry indices. Using three versions of such as Dow Jones, FTSE, MSCI and S&P, their composite index (low, medium and as well as financial institutions such as high utilities), and consistent with previous Brookfield, Macquarie or UBS. These indices papers, they report that over the 1990- are not fundamentally different from the 2007 period, infrastructure returns were approach described above. similar to that of U.S. equities but with slightly less risk. Finally, using the CAPM They use asset selection schemes based on to create a forward-looking model of slightly different industrial definitions of expected returns including an infrastructure

A Publication of the EDHEC Infrastructure Institute-Singapore 15 Searching for a listed infrastructure asset class- June 2016

2. Literature Review

allocation, Idzorek and Armstrong (2009) find that adding infrastructure does not lead to a meaningful improvement of the efficient frontier, but again provide no statistical test of the robustness of their results.

2.3 Limitations of existing research The existing literature has not examined whether different types of listed infrastructure investments are not already spanned in the portfolio of a typical investor. As a result, it remains unclear whether a focus on infrastructure-related stocks can create additional diversification benefits for investors. Nor is it clear whether ”infrastructure” is a new combination of investment factor exposures.

In the rest of this paper, we test statistically whether infrastructure stocks, selected according to their industrial classification, provide diversification benefits to investors. Furthermore, following the argument in Blanc-Brude (2013), we examine a different definition of infrastructure focussing on the ”business model” as determined by the role of long-term contracts in infrastructure projects. Next, we describe our approach in more detail.

16 A Publication of the EDHEC Infrastructure Institute-Singapore 3. Approach

A Publication of the EDHEC Infrastructure Institute-Singapore 17 Searching for a listed infrastructure asset class- June 2016

3. Approach

3.1 Testing three definitions of with the public sector to build, maintain listed infrastructure and operate public infrastructure facilities We propose to test the portfolio character- according to a pre-agreed service output istics of listed infrastructure equity under specification. As long as these firms deliver the three different definitions of what the projects and associated services for constitutes ”infrastructure” proposed in which they have been mandated, on time section 1. and according to specifications, the public sector is committed to pay a monthly or These first two proxies – a ”naïve”, rule- quarterly income to the firm according to based filtering of stocks based on industrial a pre-agreed schedule for multiple decades. sector classifications and listed infras- In the UK, the long-term contract between tructure indices maintained by index the public and private parties also stipu- providers – focus on the ”real” charac- lated that this ”availability payment” is also teristics of the relevant capital projects adjusted to reflect changes in the retail and bundles together assets that may price index (RPI). Each project company all be related to large structures of steel is a special purpose vehicle created solely and concrete, but may also have radically to deliver an infrastructure project and different risk profiles from an investment financed on a non-recourse basis with view point. sponsor equity, shareholder loans and senior debt. Hence, we also identify stocks which happen 2 The firms we identify are listed on the 2 - i.e. Experimental and control to create a useful natural experiment : a conditions are determined by factors basket of stocks offering a pure exposure to London Stock Exchange, buy and hold the outside our control projects that correspond to a specific long- equity and shareholder loans (quasi-equity) term contract but not to any specific indus- of hundreds of these PFI project companies trial sectors. and subsequently distribute dividends to they shareholders. They are, in effect, a listed Instead, this basket captures a specific basket of PFI equity (with no additional infrastructure ”business model”: these are leverage) and, as such represent a pure the publicly traded shares of investment proxy of one model of private infrastructure vehicles that are solely involved in buying investment. We provide a full list of the and holding the equity of infrastructure underlying firms in the appendix. projects engaged in PFI (Private Finance Initiative) projects in the UK and, to a lesser We expect the cash flows of these firms extent, their equivalents in Canada, France to be highly predictable, uncorrelated with and the rest of the OECD. markets and the business cycle, albeit highly correlated with the UK RPI index. In other PFI projects consists of dedicated project words, we expect to see some evidence firms entering into long-term contracts of the ”infrastructure investment narrative”

18 A Publication of the EDHEC Infrastructure Institute-Singapore Searching for a listed infrastructure asset class - June 2016

3. Approach

discussed earlier, that has so far eluded assuming that assets belong to distinctive studies of infrastructure stocks defined by categories because they have different their SIC code. names. 3 3 - Such factors include the Fama and French (1992) Size and Value premiums, the Term and Default premiums (Fama and French, 1993) Thus, we include both a traditional asset and the Momentum premium identified by Jegadeesh and Titman 3.2 Testing asset classes or allocation based on asset classes and a (1993). Bender et al. (2010) shows that these premiums are uncorrelated factors? factor based allocation to test whether with each other, increase returns listed infrastructure is indeed an asset class and reduce portfolio volatility over Using these three alternative approaches to traditional asset class allocations. or, alternatively, a unique combination of Likewise, when comparing the diver- define infrastructure investment, we use the sification benefits of factor-based mean-variance spanning tests designed by investment factors. allocations to alternative assets, Bird et al. (2013) finds that factor Huberman and Kandel (1987) and Kan and approaches tend to outperform alternative asset classes. For recent Zhou (2012) to determine whether adding and in-depth analyses of factor investing, see Amenc et al. (2014). a listed infrastructure bucket to an existing 3.3 Testing persistence investment portfolio significantly increases Finally, we test the existence of a persistent diversification opportunities. effect of listed infrastructure on a reference portfolio by splitting the observation period In the affirmative, this result implies a in two, from 2000 to 2008 and from 2008 degree of ”asset class-ness” of infras- to 2014, to test for the impact of the 2008 tructure stocks since their addition to a reversal of the credit cycle a.k.a. the Global reference portfolio effectively shifts the Financial Crisis (GFC) and the distorting role mean-variance efficient frontier (to the left) excessive leverage may have played in the and creates new diversification opportu- first period. In section 6, we report results nities for investors. Furthermore, we define for the whole sample period, as well as for the reference portfolio used to test the two sub-sample periods denoted as pre- and mean-variance spanning properties of listed post-GFC periods. infrastructure either in terms of traditional asset classes and of investment factors. In the case of the PFI portfolio, data starts in 2006 and so we also split the sample in Indeed, the notion of asset class is losing its 2011, which marks the time of the Eurozone relevance in investment management since debt crisis and launch of quantitative easing the financial crisis of 2008-11,when existing policies by the Bank of England. asset class-based allocations failed to prove well diversified (see for example Ilmanen In the next section, we discuss the mean- and Kizer, 2012). variance spanning methodology used in the remaining sections of this paper. Factor-based asset allocations aim to identify those persistent dimensions of financial assets that best explain (and predict) their performance instead of

A Publication of the EDHEC Infrastructure Institute-Singapore 19 4. Methodology

20 A Publication of the EDHEC Infrastructure Institute-Singapore Searching for a listed infrastructure asset class - June 2016

4. Methodology

In a mean-variance framework, the been used in the literature on listed infras- question of whether listed infrastructure tructure. provides diversification benefits is equiv- alent to asking whether investors are able Mean-variance spanning is a regression to improve their optimal mean-variance based test that assumes that there are K frontier by including infrastructure stocks reference assets as well as N test assets. to an existing portfolio. In Huberman and Kandel (1987), there is a linear relationship between test and This question can be answered using the reference assets so that: mean-variance spanning test described by = α + β + ε Huberman and Kandel (1987), which tests R2t R1t t (4.1) whether the efficient frontier is improved with t = 1,... T periods and R1 represents up to given level of statistical signifi- a T × K matrix of excess realised returns cance when including new assets. of the K benchmark assets. R2 represents a T × N matrix of realised excess returns of If the mean-variance frontier, inclusive of the N test assets. β is a K × N matrix of the new assets, coincides with that already regression factor loadings and ε is a vector produced by the reference assets, the new of the regression error terms. assets can be considered to be already spanned by the existing portfolio i.e. no new The null hypothesis is that existing assets diversification benefit is created. Conversely, already span the new assets. This implies if the existing mean-variance frontier is that the α of the regression in equation 4.1 shifted to the left in the mean/variance is equal to zero, whilst the sum of the βs plane, by the addition of the new asset, equals one. As a result, the null hypothesis investors have improved their investment assumes that a combination of the existing opportunity set. benchmark assets is capable of replicating the returns of the test assets with a lower This approach has been used to examine variance. the diversification benefits in different asset classes. For instance, Petrella (2005) and Eun Kan and Zhou (2012) describes the null et al. (2008) employ this methodology to hypothesis as : examine the diversification benefits of small − cap stocks. Likewise, Kroencke and Schindler H0S = α = 0N, δ = 1N β1K = 0N (2012) examines the benefits of interna- (4.2) tional diversification in real estate using mean-variance spanning, while Chen et al. Where αN is an N-vector of regression (2011) examines the diversification benefits intercept coefficients and β is the matrix of of volatility. However, to date it has not factor loadings.

A Publication of the EDHEC Infrastructure Institute-Singapore 21 Searching for a listed infrastructure asset class- June 2016

4. Methodology

As this analysis is only examining the The second stage of the Kan and Zhou

addition of one test asset at a time the test (2012) test examines whether δ = 0N statistic is given by: 4 conditional on α = 0 . If the null 4 - Kan and Zhou (2012) state that N ≥ ( )( ) if N 2 then the appropriate − − hypothesis is rejected, the Global Minimum formation( of the test) statistic( is given) 1 − T K 1 − − HK = 1 (4.3) as HK = 1 − 1 T K 1 . Variance of the test portfolio and the 1 2 V 2 V 2 benchmark portfolios are statistically different (see also Chen et al., 2011, for a where V is the ratio of the determinant discussion). of the maximum likelihood estimator of the error co-variance matrix of the model In this paper, we incorporate both the assuming that there is no spanning of Huberman and Kandel (1987) and Kan and the efficient frontier (otherwise known as Zhou (2012) tests to examine the ability of the unrestricted model) to that of the infrastructure to provide portfolio diversifi- determinant of the maximum likelihood 5 5 - As another robustness check we cation benefits. employ the Gibbons et al. (1989) test estimator of the model that assumes of portfolio efficiency. The results, are similar to the mean-variance span spanning occurs (known as the restricted Next, we describe the data used in this study. test results presented in this paper. model). The results are available upon request.

T is the number of return observations, K is the number of benchmark assets included in the study. The HK variable is a Wald-test statistic and follows an F-distribution with (2, T − K − 1) degrees of freedom.

In addition to the Huberman and Kandel (1987) test, Kan and Zhou (2012) develop a two-stage test to examine whether the rejection of the Huberman and Kandel (1987) null hypothesis is due to differences in the tangency or the Global Minimum Variance as a result of the addition of new assets.

The first step of the Kan and Zhou (2012)

test examines whether α = 0N. If the null is rejected at this stage, the two tangency portfolios comprising of the benchmark assets, and the benchmark and new assets, respectively, are statistically different.

22 A Publication of the EDHEC Infrastructure Institute-Singapore 5. Data

A Publication of the EDHEC Infrastructure Institute-Singapore 23 Searching for a listed infrastructure asset class- June 2016

5. Data

This section describes the datasets used to is a crude measure as it relies on the build test portfolios of listed infrastructure continuous updating of the revenue codes and reference portfolios to apply the mean- by Worldscope, as well as assuming that variance spanning methodology described GIC or SIC codes represent infrastructure previously. activities.

Sections 5.1, 5.2, and 5.3 describe listed Setting a minimum infrastructure sector infrastructure proxies designed with sector- revenue threshold to 75% and 90%, yields based asset selection rules, index-provider 650 and 554 stocks, respectively. data, and the PFI portfolio, respectively. Section 5.4 describes the reference U.S. dollar price and total returns are portfolios. sourced from Datastream using the methodology described in Ince and Porter (2006). 7 The firms thus identified comprise 5.1 Test Assets - Listed at most 12%, 7% and 6.5% of the of infrastructure companies the MSCI World market value as at 31 5.1.1 Asset selection December 2014, for the 50%, 75% and The first asset selection scheme repre- 90% infrastructure revenue thresholds, sents the ”naïve” definition of infrastructure respectively. equity investment, and follows the methodology described by Rothballer and Kaserer 5.1.2 Descriptive Statistics (2012b) following broad industry definitions For market-cap weighted portfolios of to determine infrastructure-related stocks 6. infrastructure stocks defined according 6 - The SIC and GIC codes used to identify infrastructure are available to the industry-based scheme described upon request. above, we report annualised returns, 7 - Extreme monthly returns are 5,757 possible securities listed in identified following Ince and Porter standard deviation and Sharpe ratios and (2006) and set to a missing value. global markets are thus identified as Ince and Porter (2006) sets an arbitary infrastructure-related. Next, only stocks the maximum drawdown ratios for the cut off of 300% for extreme monthly returns. If R1 or Rt−1 is greater than for which the majority of the revenue was period 2000-2014 for price and total − 300% and (1 + R1)/(1 + Rt−1) 1 returns. in table 1. is less than 50% then R1 or Rt−1 are obtained from sectors corresponding to set to missing. Furthermore, following Rothballer and Kaserer (2012b), 18 infrastructure activities are kept in the months of non zero returns are It is useful to note that the reference required for the stock to be included sample. in the portfolios. market index should not be difficult to Any Datastream padded price is removed by requesting X(P#S) $U A minimum market capitalisation of beat. The MSCI World index is a free float which returns null values when Datastream does not have a record USD500 Million is also required to be adjusted market capitalisation weighted and any non equity item is removed index comprising of 1,631 mid and large by requiring the TYPE description in included in the sample. This yields 1,290 Datastream to be equal to EQ. firms with at least 50% of their income capitalisation stocks across 23 developed from infrastructure activities The minimum country equity markets. MSCI states revenue by infrastructure type is reported that the index comprises 85% of the by SIC or GIC code by Worldscope. This free-float adjusted market capitalisation

24 A Publication of the EDHEC Infrastructure Institute-Singapore Searching for a listed infrastructure asset class - June 2016

5. Data

Table 1: This table presents the descriptive statistics for the naïve infrastructure stock selection scheme, 2000-2014. Telecom, Transport and Utilities are portfolios of stocks that earn a minimum revenue level from activities related to SIC or GIC codes recognised as Telecommunications, Transport and Utilities industries, respectfully. The minimum revenue levels required are 50%, 75% and 90% , respectfully. Return is the average monthly return from January 2000 to December 2014. Risk is the monthly standard deviation of returns from January 2000 to December 2014. SR is the Sharpe Ratio calculated with the average yield of the U.S. one month Treasury bill as the risk free rate proxy. Worst Drawdown is the maximum drawdown ratio, measured as a percentage of maximum cumulative return i.e. from ”peak equity.

Price returns , 50% Rev Tel. 50% Rev Transp. 50% Rev Util. MSCI World Price return −0.084 0.020 −0.005 0.012 Risk 0.179 0.154 0.152 0.158 SR −0.496 0.094 −0.065 0.047 Worst Drawdown 0.830 0.620 0.570 0.550 , 75% Rev Tel. 75% Rev Transp. 75% Rev Util. MSCI World Price return −0.092 0.068 0.003 0.012 Risk 0.185 0.198 0.133 0.158 SR −0.522 0.318 −0.015 0.047 Worst Drawdown 0.830 0.690 0.500 0.550 , 90% Rev Tel. 90% Rev Transp. 90% Rev Util. MSCI World Price return −0.085 0.042 0.002 0.012 Risk 0.175 0.180 0.133 0.158 SR −0.515 0.205 −0.025 0.047 Worst Drawdown 0.810 0.690 0.480 0.550 Total returns , 50% Rev Tel. 50% Rev Transp. 50% Rev Util. MSCI World Tot. return −0.052 0.048 0.028 0.036 Risk 0.180 0.152 0.153 0.159 SR −0.315 0.282 0.148 0.197 Worst Drawdown 0.818 0.609 0.556 0.537 , 75% Rev Tel. 75% Rev Transp. 75% Rev Util. MSCI World Tot. return −0.057 0.109 0.039 0.036 Risk 0.185 0.197 0.134 0.159 SR −0.334 0.523 0.255 0.197 Worst Drawdown 0.826 0.667 0.473 0.537 , 90% Rev Tel. 90% Rev Transp. 90% Rev Util. MSCI World Tot. return −0.051 0.088 0.038 0.036 Risk 0.176 0.179 0.133 0.159 SR −0.315 0.461 0.249 0.197 Worst Drawdown 0.799 0.662 0.454 0.537

A Publication of the EDHEC Infrastructure Institute-Singapore 25 Searching for a listed infrastructure asset class- June 2016

5. Data

of each country covered. The index is 5.2 Test assets - Ad hoc listed updated quarterly with annual revisions infrastructure indices to update the investable universe and the 5.2.1 Asset selection removal of stocks with low liquidity. Such The basic requirements to be included in market value-weighted indices, while they listed infrastructure indices created by index constitute a useful point of reference, have providers are not very different from the been shown to be highly inefficient in naïve selection scheme described above. previous research (see Amenc et al., 2010). They include:

Nevertheless, the listed infrastructure 1. being part of a broader index universe portfolios obtained above do not necessarily (usually that of the infrastructure offer better risk-adjusted performance than universe of the index provider); and, this relatively unambitious baseline. 2. a minimum amount of revenue derived from infrastructure activities. We observe that irrespective of the revenue cut-offs employed to form the infras- However, minimum revenue requirements tructure portfolios, the telecom sector and the definition of infrastructure activ- continually produces poor returns. This ities are set differently by each index sector does not seem to have recovered provider, adding what could amount to from the technology bubble of the yearly ”active views”, to a rule-based scheme. 2000s. This suggests that certain ”infrastructure” sectors experience a high degree We test two groups of listed infrastructure of cyclicality, as well as a complete absence indices: a set of global indices and one of persistence. Transportation fares better designed to represent the U.S. market only. with higher Sharpe ratios than the market Global indices provide a direct comparison index under both the price return and with the naïve approach described above, total return measures. However, drawdown while a U.S.-only perspective allows more risk is typically higher than the market’s controls and granularity when designing a for price and total returns as shown in reference portfolio of asset classes or factors Table 1. During the sample period Utilities to test the mean-variance spanning of listed only outperform the broad market from a infrastructure indices. total return perspective. Global Infrastructure Indices Next, we describe our second test asset, We include seven global infrastructure a combination of rule-based and ad hoc 8 8 - A brief summary of the indices is indices and four U.S. infrastructure indices: available upon request stock selection schemes created by index providers. G Dow Jones Brookfield Global Infras- tructure Index;

26 A Publication of the EDHEC Infrastructure Institute-Singapore Searching for a listed infrastructure asset class - June 2016

5. Data

G FTSE Macquarie Global Infrastructure index which exhibits negative performance Index; on a price return basis. However Table 2 G FTSE Global Core Infrastructure; suggest that drawdown risk is very similar G MSCI World Infrastructure Index; between infrastructure indices and the G MSCI ACWI Infrastructure Capped; broad market, with the exception of the G UBS Global Infrastructure and Utilities; Brookfield and MSCI ACWI indices. and, G UBS Global 50/50 Infrastructure and U.S. infrastructure indices Utilities.

The universe thus recognised by index In the case of US-only indices, the MSCI providers is not very large with only the and FTSE indices, reported in Table 3 do MSCI World Infrastructure and MSCI ACWI not seem to stand out from the broad Global Infrastructure representing more market index (here the Russell 3000), but the than 10% of the value of the MSCI World Alerian MLP index, which captures a under- Index. lying business model focused on dividend distributions, exhibits very different charac- U.S. infrastructure indices teristics. with much higher Sharpe ratios The U.S. infrastructure indices included in especially on a total return basis, but equally this study are: high maximum drawdown.

G FTSE Macquarie USA Infrastructure Index; G MSCI US Infrastructure Index; 5.3 Test Assets - Listed baskets of G MSCI USA Infrastructure 20/35 Capped contracted infrastructure projects Index; and, 5.3.1 Asset selection G Alerian MLP Infrastructure Index. The PFI portfolio consists of

5.2.2 Descriptive statistics 1. HSBC Infrastructure Company Ltd (HICL) Global Infrastructure Indices 2. John Laing Infrastructure Fund Ltd (JLIF) 3. GCP Infrastructure Ltd (GCP) 4. International Partnerships Ltd (INPP) Table 2 shows that most infrastructure 5. Bilfinger Berger Global Infrastructure Ltd indices have higher Sharpe ratios than (BBGI) the reference market index (MSCI World). The Dow Jones Brookfield Global Infras- As discussed, these firms are solely tructure index exhibits the highest average occupied with buying and holding the annualised returns and Sharpe ratio for equity and quasi-equity of PFI (private the sample period. This performance is finance initiative) project companies in contrasted by the MSCI World Infrastructure existence in the U.K. and that of similar

A Publication of the EDHEC Infrastructure Institute-Singapore 27 Searching for a listed infrastructure asset class- June 2016

5. Data

Table 2: This table presents the descriptive statistics for the Global Infrastructure Indices for the period, 2000-2014. BF is the Dow Jones Brookfield Global Infrastructure Index, SPis Standard & Poor’s Global Infrastructure Index, FTSEM is the FTSE Macquarie Global Infras- tructure Index, FTSEC is the FTSE Global Core Infrastructure Index, MSCI is the MSCI World Infrastructure Index, MSCIA is the MSCI ACWI Infrastructure Capped, UBS is the UBS Global Infrastructure and Utilities, UBS 50 is the UBS Global 50/50 Infrastructure and Utilities Index and MSCIW is the MSCI World Index. Return is the average monthly return from the index commencement date to December 2014. Risk is the monthly standard deviation of returns from the index commencement date to December 2014. SR is the Sharpe Ratio calculated with the average yield of the U.S. one month Treasury bill as the risk free rate proxy. Worst Drawdown is the maximum drawdown ratio, measured as a percentage of maximum cumulative return i.e. from ”peak equity”.

Price returns , BF SP FTSEM FTSEC MSCI MSCIA UBS UBS50 MSCIW Return 0.112 0.072 0.043 0.063 −0.020 0.025 0.046 0.055 0.012 Risk 0.132 0.153 0.138 0.115 0.145 0.105 0.119 0.145 0.158 SR 0.807 0.436 0.273 0.507 −0.171 0.192 0.347 0.341 0.047 Worst Drawdown 0.476 0.551 0.456 0.374 0.660 0.424 0.447 0.505 0.554 Total returns , BF SP FTSEM FTSEC MSCI MSCIA UBS UBS50 MSCIW Return 0.147 0.116 0.083 0.099 0.019 0.061 0.082 0.091 0.036 Risk 0.132 0.153 0.138 0.114 0.146 0.105 0.119 0.145 0.159 SR 1.070 0.725 0.561 0.820 0.092 0.532 0.644 0.588 0.197 Worst Drawdown 0.452 0.527 0.432 0.348 0.640 0.395 0.426 0.484 0.537

Table 3: This table presents the descriptive statistics for of annualised price and total returns of U.S. infrastructure stock indices, 2000-2014. AMLP is the Alerian MLP Infrastructure Index, FTSEM is the FTSE Macquarie USA Infrastructure Index, MSCI is the MSCI US Infrastructure Index, MSCISC is the MSCI USA Infrastructure 20/35 Capped Index and R3000 is the Russell 3000 index. Return is the average monthly return from January 2000 to December 2014. Risk is the monthly standard deviation of returns from January 2000 to December 2014. SR is the Sharpe Ratio calculated with the average yield of the U.S. one month Treasury bill as the risk free rate proxy. Worst Drawdown is the maximum drawdown ratio, measured as a percentage of maximum cumulative return i.e. from ”peak equity”.

Price returns , AMLP FTSEM MSCI MSCISC R3000 Price return 0.130 0.063 −0.016 0.024 0.029 Risk 0.166 0.448 0.147 0.138 0.157 SR 0.748 0.130 −0.141 0.133 0.155 Worst Drawdown 0.492 0.956 0.633 0.448 0.527 Total returns , AMLP FTSEM MSCI MSCISC R3000 ann. total return 0.213 0.067 0.021 0.055 0.048 ann. risk 0.170 0.448 0.148 0.138 0.157 ann. Sharpe ratio 1.219 0.137 0.106 0.361 0.274 Worst Drawdown 0.431 0.956 0.609 0.423 0.512

28 A Publication of the EDHEC Infrastructure Institute-Singapore Searching for a listed infrastructure asset class - June 2016

5. Data

firms mostly involved in delivering so- than the market reference (here the FTSE All called availability-payment infrastructure Shares). Indeed, the maximum drawdown projects, by which the public sector pays for the PFI portfolio is also much lower a pre-agreed income to the project firm than the FTSE Macquarie Europe infras- on a regular basis in exchange for the tructure index, another listed infrastructure construction/development, maintenance index focused solely on European markets. and operations of a given infrastructure The combination of high risk-adjusted project given a pre-agreed output specifi- performance with low drawdown risk is cation and for several decades. particularly striking in the total return case.

These PFI project companies in question do not enter into any other activities 5.4 Reference Assets during their lifetime, and solely deliver the As discussed above, we use two types of contracted infrastructure and associated reference allocations to test the impact of services while repaying their creditors and adding listed infrastructure to an investor’s investors. As such, they give access to universe, an asset class-based allocation and a “pure” infrastructure project cash flow, a factor-based allocation. All the summary representative of the underlying nature of statistics for the reference assets can be the PFI business model. found in table 13 in the Appendix.

The firms in the PFI portfolio can be 5.4.1 Global asset class-based considered useful proxies of a portfolio of reference portfolio PFI equity investments. While the project A ”well diversified investor” in the traditional companies are typically highly leveraged, albeit imprecise meaning of the term can be the firms in the PFI portfolio do not make expected to hold a number of different asset a significant use of leverage. Hence, asa classes, including: group, they can be considered to be representative a listed basket of PFI equity stakes. G Global Fixed Interest proxied by JP Morgan Global Aggregate Bond Index; All returns are annualised monthly price and G Commodities proxied by The S&P total returns computed in local currency Goldman Sachs Commodity Index; (GBP) and sourced from Datastream. G Real Estate proxied by MSCI World Real Estate Index; 5.3.2 Descriptive Statistics G Hedge Funds proxied by Dow Jones Credit Table 4 suggest that the PFI portfolio Suisse Hedge Fund Index; and, possesses different characteristics than G OECD and Emerging Market Equities other listed infrastructure portfolios proxied by MSCI World and MSCI examined so far. Its Sharpe ratio is high Emerging Market Indices, respectively. but its maximum drawdown is much lower

A Publication of the EDHEC Infrastructure Institute-Singapore 29 Searching for a listed infrastructure asset class- June 2016

5. Data

Table 4: This table presents the descriptive statistics for of annualised price and total returns of the PFI portfolio, an infrastructure index and the market index, 2006-2014. The PFI Portfolio is the equal-weighted return of the PFI stocks identified listed on the London Stock Exchange. Return is the average monthly return from 2006 to December 2014. Risk is the monthly standard deviation of returns from 2006 to December 2014. SR is the Sharpe Ratio calculated with the average yield of the U.S. one month Treasury bill as the risk free rate proxy. Max. DD is the maximum drawdown ratio, measured as a percentage of maximum cumulative return i.e. from ”peak equity”. All returns are annualised monthly price and total returns computed in local currency (GBP) and sourced from Datastream.

Price returns , PFI Portfolio FTSE All Shares Macquarie Infra Europe Price return 0.048 0.027 −0.007 Risk 0.093 0.182 0.181 SR 0.460 0.121 −0.065 Max. DD 0.240 0.450 0.500 Total returns , PFI Portfolio FTSE All Shares Macquarie Infra Europe Tot. return 0.101 0.065 0.046 Risk 0.082 0.172 0.184 SR 1.171 0.345 0.224 Max. DD 0.150 0.410 0.370

One potential issue with employing indices we cannot know the constituents and so as a reference asset is the possibility of this cannot be done. double-counting infrastructure stocks in both the reference and test assets. This has Nevertheless, given the low weighting the potential of biasing the mean-variance its likely that any results will not be span tests against finding an improvement biased against the infrastructure stocks. We in the investment opportunity set. Whilst conclude that not isolating infrastructure ideally removing any infrastructure like stocks from our reference assets will not stocks from the reference assets would solve materially influence the conclusions of this the problem of double counting, the circu- study. lation of index constituent lists is too limited to allow this. 5.4.2 U.S. asset class reference portfolio However, the MSCI World Index, MSCI A typical U.S.-based reference portfolio (2014) states that as at November 2014 the built using traditional asset classes would Utilities and Telecom industries comprise include: 3.32% and 3.46% whilst the Industrials sector comprises 10.89% and the share of G Government Bonds proxied by the ”infrastructure” in industrials (e.g. railway) is Barclays Govt Aggregate Index; small. G Corporate Bonds represented by the Barclays U.S. Aggregate Index; Whilst it would be preferable to exclude the G High Yield Bonds with the Barclays U.S. infrastructure stocks from the MSCI World, Corporate High Yield;

30 A Publication of the EDHEC Infrastructure Institute-Singapore Searching for a listed infrastructure asset class - June 2016

5. Data

G Real Estate, as per the US-DataStream G The Market factor is the excess return of Real Estate Index; the MSCI U.S. and MSCI Europe indices. G Hedge Funds represented by the Dow G The Size factor (SMB) is calculated by Jones Credit Suisse Hedge Fund Index; taking the difference between the simple G Commodities proxied by the S&P average of MSCI Small Value and Growth Goldman Sachs Commodity Index; indices and the simple average of MSCI G U.S. Equities captured by the RUSSELL Large Value and Growth Indices. 9 G 9 - The Russell 3000 index was 3000 ; and, The Value factor (HML) is constructed by selected for the U.S. equity market G World Equities represented by the MSCI obtaining the difference between simple index for two reasons. Firstly, it represents the top 3,000 stocks by market World ex-US. average of MSCI Small, Mid and Large capitalisation. This represents a significant proportion of the investable Value indices and simple average of MSCI universe of U.S. stocks. Secondly, for consistency, in the factor exposure 5.4.3 U.K. asset class reference Small, Mid and Large Growth Indices. studies we employ the Russell indices portfolio G The Term factor is estimated by taking the to create the factor proxies. To test the mean variance spanning difference between the returns of the U.S. properties of the PFI portfolio, we build Government 10 year index and S&P U.S. a U.K. asset class reference portfolio Treasury Bill 0-3 Index. consisting of: G Finally, the Default factor is estimated by the change in the Moody’s Seasoned Baa G Fixed Interest, represented by the Bank of Corporate Bond Yield Relative to the Yield America/ML U.K. Gilts index; on 10-Year Treasury Constant Maturity. G Real Estate, proxied by the DataStream U.K. Real estate Index; 5.4.5 U.S. factor-based reference G Hedge funds, represented by the U.K. portfolio DataStream hedge funds Index; U.S. factors are computed using the now canonical formulas reported in (Faff, 2001): G Commodities, as proxied by the S&P Goldman Sachs Commodity Index; Market = Russell 3000 Index −US one month Treasury Bill return G U.K. Equities represented by the FTSE100; and. (Russell2000Value + Russell2000Growth) SMB = G World Equities proxied by the FTSE World 2 ex-U.K. − (Russell1000Value + Russell1000Growth) 2

5.4.4 Global factor-based reference (Russell1000Value + Russell2000Value) HML = portfolio 2 − (Russell1000Growth + Russell2000Growth) Consistent with prior research, the factors 2 in this study are constructed from stock and = bond market indices. We follow Bender et al. Term Barclays US Treasury 10-20 years Index −Barclays US Treasury Bills 1-3 months Index (2010), Ilmanen and Kizer (2012) and Bird et al. (2013) to build Market, Size, Value, Default = Barclays US Corporate: AAA Long Index Term and Default factors. −Barclays US Treasury Long Index

A Publication of the EDHEC Infrastructure Institute-Singapore 31 Searching for a listed infrastructure asset class- June 2016

5. Data

In the next section, we present the results of the mean-variance spanning tests presented in section 4 using the multiple datasets described above.

32 A Publication of the EDHEC Infrastructure Institute-Singapore 6. Results

A Publication of the EDHEC Infrastructure Institute-Singapore 33 Searching for a listed infrastructure asset class- June 2016

6. Results

We first present the results of the Mean- investments. Indeed, the listed infras- Variance Spanning tests conducted using tructure portfolios fail to improve either the asset classes as the reference assets in the tangent portfolio or produce a lower section 6.1. Next, we use the factors defined minimum variance portfolio. This finding is above as the reference portfolio. consistent when either price or total returns are used. We report the Huberman and Kandel (1987) regression results for equation 4.1 and the When the sub-periods are considered, both corresponding Kan and Zhou (2012) two before and after the GFC, the conclusion stage, step down tests. that the naïve infrastructure approach fails to identify any diversification benefits is supported. 6.1 Asset class mean-variance spanning test results Panels B and D in Table 5 present the results 6.1.1 Listed infrastructure companies for the Mean-Variance Spanning tests for The results of the mean-variance span test the period January 2000 to December 2008. for the naïve infrastructure portfolios are The Huberman and Kandel (1987) test’s null reported in Table 5. For the price returns hypothesis is rejected in two cases: the of the nine portfolios constructed, Table 5 price and total returns of the 75% Transport shows that the reference investment portfolio. However, the Kan and Zhou (2012) opportunity set is improved by four of these test fails to reject the null hypothesis that portfolios. These are the 75% Revenue the reference portfolio already spans Cutoff Transport portfolio as well as the listed infrastructure. Telecommunication portfolios. However, when total returns are examined, only From January 2009 to December 2014 (Panel the 50% Telecoms and the 75% Transport C and F in Table 5) only one portfolio is found infrastructure portfolio are found to to improve the efficient frontier, the price reject the Huberman and Kandel (1987) returns of the 50% Utilities portfolio. Here, null hypothesis of spanning at conven- the Huberman and Kandel (1987) portfolio’s tional significance levels. Every other null hypothesis is rejected at the 5% signif- portfolio does not improve upon the icance level and both steps of the Kan and mean-variance frontier created by the Zhou (2012) test reject the null hypothesis reference asset classes. that the portfolio’s risk and returns are already spanned by the reference assets. Applying the Kan and Zhou (2012) methodology, the results in Table 5 shows that This seldom provides systematic evidence of none of the infrastructure portfolios the existence of a listed infrastructure asset improve the mean-variance frontier class. from that created by the reference

34 A Publication of the EDHEC Infrastructure Institute-Singapore Searching for a listed infrastructure asset class - June 2016

6. Results

Figure 3: Mean-variance frontier of 90% revenue threshold utilities and asset class reference portfolio

0.20 Efficient Frontier January 2009 to December 2014

● Real Estate 0.15

● Developed Equities ● Emerging Equities

0.10

● Hedge Funds Returns

0.05 ● Fixed Interest ●90% Utilities

0.00

● Commodities

−0.05 0.00 0.05 0.10 0.15 0.20 0.25 0.30 Standard Deviation

Reference Assets Reference and Infrastructure Assets

An illustration of the findings in table 5is with a global asset class-based reference shown in figure 3, which presents the mean- portfolio. variance frontier with and without the addition of the naïve 90% Utilities portfolio Global infrastructure indices for the period January 2009 to December Table 6 presents our results for the global 2014. The results in table 5 confirm that this infrastructure price and total return indices portfolio does not improves the investment described in section 5. Here, using price opportunity set despite shifting the efficient returns for the full sample period (Panel frontier to the left since the minimum A, Table 6), the Huberman and Kandel variance point is not statistically different (1987) test finds an improvement in the before and after adding ”infrastructure” to efficient frontier in six of the eight infras- the asset mix. tructure indices examined. However, the more restrictive Kan and Zhou (2012) test Next, we discuss our results using industry- only finds two of the eight global infras- provided infrastructure indices as proxies tructure indices to improve the reference of the infrastructure asset class, testing efficient frontier: the Dow Jones Brookfield whether there are diversification benefits Global Infrastructure index and the UBS Infrastructure and Utilities index.

A Publication of the EDHEC Infrastructure Institute-Singapore 35 Searching for a listed infrastructure asset class- June 2016

6. Results

Other indices found to improve the efficient the minimum variance portfolio, and not frontier by the Huberman and Kandel (1987) the tangency portfolio. test are instead found to improve the tangency portfolio or the minimum variance When total returns are considered for the portfolio but not both. As a result, it cannot same period, the Huberman and Kandel be accepted that these indices improve the (1987) test finds all of the listed infras- efficient frontier. tructure indices improve the efficient frontier. The results of the Kan and Zhou Using total returns (Panel D, table 6), again (2012) test however, indicate that while six of the eight global indices reject the during this period the global indices null of the Huberman and Kandel (1987) improved on the tangency portfolio, not all test. Only the FTSE Core index fails to span impacted the minimum variance portfolio. when either the price or total returns are As a result, pre-GFC, only FTSE Core, MSCI employed, while the MSCI World Infras- World Infrastructure and the MSCI ACWI tructure is not spanned by the reference Capped infrastructure indices can be said asset classes using price returns but is when to improve the efficient frontier, as they considering total returns. The reverse is true both improve the tangency portfolio and for the UBS 50-50. reduce the minimum variance portfolio for this period. Using the Kan and Zhou (2012) test, four of the eight global infrastructure indices are Post-GFC (panels E and F of table 6), pre-GFC found to improve the efficient frontier. But results are invalidated. Using price returns, table 6 shows that most indices found by only one of the eight indices examined the Huberman and Kandel (1987) test to is found to improve the efficient frontier improve the efficient frontier only improved under both the Huberman and Kandel the minimum variance portfolio, but not (1987) and Kan and Zhou (2012) tests: the improve the tangency portfolio. As a result, Dow Jones Brookfield index. Using total it is not possible to conclude that these returns again only one index is found to listed infrastructure indices are not spanned improve the efficient frontier under both by existing asset classes. the Huberman and Kandel (1987) and Kan and Zhou (2012) tests: the MSCI ACWI Turning to sub-periods, for price returns Capped index. pre-GFC (Panels B and C of table 6), the Huberman and Kandel (1987) test Hence, pre-GFC results are not persistent finds that four of the eight global listed post-GFC. These results argue against the infrastructure indices improve the efficient existence of a well-defined and persistent frontier. However, the Kan and Zhou (2012) listed infrastructure ”asset class”. test finds that these indices only improve

36 A Publication of the EDHEC Infrastructure Institute-Singapore Searching for a listed infrastructure asset class - June 2016

6. Results

Figure 4: Illustration of the difference tests of mean-variance spanning of the FTSE Macquarie USA Infrastructure Index

(a) 2000-2014, asset class and factor-based reference Efficient Frontier January 2000 to December 2013 0.20 0.20 Efficient Frontier February 2007 to December 2013

0.15 0.15

0.10 0.10 High Yield ● ● ● US Equities ● FTSE Macquarie US Infra FTSE Macquarie US Infra

● Market

Returns ●

Real Estate Returns ● Term Corporate Bonds ● 0.05 ● Hedge Funds 0.05 ● Government Bonds ● World Ex−US Equities

● Default 0.00 0.00 ● Size

● Value

● Commodities −0.05 −0.05 0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.00 0.05 0.10 0.15 0.20 0.25 0.30 Standard Deviation Standard Deviation

Reference Assets Reference and Infrastructure Assets Reference Assets Reference and Infrastructure Assets

(b) 2000-2008, asset class and factor-based reference Efficient Frontier February 2007 to December 2008 0.25 0.20 Efficient Frontier January 2000 to December 2008

0.20

0.15 0.15

● Government Bonds 0.10 ● Default 0.10

● Corporate Bonds ● Value 0.05 ● FTSE Macq

0.00 0.05 ● Size ● Term

−0.05 ● Hedge Funds Returns Returns

−0.10 0.00 ● FTSE Macquarie US Infra ● Commodities ● High Yield −0.15 ● Europe Stocks −0.05 −0.20 ● US Equities ● World Ex−US Equities ● US Stocks

−0.25

−0.30

● Real Estate 0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 0.00 0.05 0.10 0.15 0.20 0.25 0.30 Standard Deviation Standard Deviation

Reference Assets Reference and Infrastructure Assets Reference Assets Reference and Infrastructure Assets

0.25

● US Stocks 0.15 0.20 ● Real Estate

● US Equities

● High Yield 0.15 ● FTSE Macquarie US Infra ● Europe Stocks 0.10 ● FTSE Macq

● World Ex−US Equities Returns 0.10 Returns

● Hedge Funds 0.05 ● Size ● Term

0.05 ● Corporate Bonds

● Government Bonds

0.00

● Value 0.00

● Commodities

−0.05 −0.05 0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.00 0.05 0.10 0.15 0.20 0.25 0.30 Standard Deviation Standard Deviation

Reference Assets Reference and Infrastructure Assets Reference Assets Reference and Infrastructure Assets

A Publication of the EDHEC Infrastructure Institute-Singapore 37 Searching for a listed infrastructure asset class- June 2016

6. Results

6.1.2 U.S. Infrastructure Indices hypothesis of the Huberman and Kandel The results for U.S. listed infrastructure (1987) test, as illustrated by figure 5. indices, are presented in table 7. When the Kan and Zhou (2012) test is The Huberman and Kandel (1987) results employed, none of the indices can reject in table 7 indicate that for the full period both steps of the test. It is not possible to (Panel A and D) both the price returns and conclude that the inclusion of infrastructure total returns of the Alerian Infrastructure indices improves the mean-variance of MLP index improves on the efficient frontier. traditional asset classes in that period. None of the other infrastructure indices reject the null hypothesis that the existing Using the second sub-sample period, asset class investments span the risk and none of the indices, either using total returns provided by the listed infrastructure or price returns, are found to reject the indices. Figure 4 provides an illustration. null hypothesis leading to the conclusion that none of the indices represent an When the Kan and Zhou (2012) test is improvement an investor’s diversification employed, the conclusion that the Alerian opportunities. Infrastructure MLP index improves the investment opportunity set is reversed for 6.1.3 PFI portfolio both the price and total return indices. Finally, we report the ability of our PFI Whilst the Kan and Zhou (2012) finds that portfolio to improve the mean-variance the tangency portfolio has improved, it efficiency of a diversified investor inthe does not reject the null hypothesis that the United Kingdom in table 8. For the complete global minimum variance has improved. As a sample, the price return series does not result, it is not possible to conclude that the provide diversification benefits. However, inclusion of the Alerian Infrastructure MLP total return results are found to improve index improves the investment universe. As on the reference efficient frontier when the other infrastructure indices do not reject investing over the entire period, as figure 6 the null hypothesis, the same conclusions illustrates. The total returns PFI portfolio apply. passes both the Huberman and Kandel (1987) and the Kan and Zhou (2012) tests When pre- and post-GFC sub-samples are for the full sample period. considered (Panels B, C, E and F), the conclusion that listed infrastructure assets Looking at sub-periods in panels, diversifi- don’t improve the investment universe, is cation benefits appear only in the period still supported. For the first sub-period, following the GFC. Prior to the GFC, neither only the total returns of the Alerian price nor total returns of the PFI portfolios Infrastructure MLP index rejects the null improve the efficient frontier. Total returns for example produce diversification benefits

38 A Publication of the EDHEC Infrastructure Institute-Singapore Searching for a listed infrastructure asset class - June 2016

6. Results

Figure 5: Mean-Variance Frontier of Alerian MLP Index Asset Class Proxies

Efficient Frontier January 2000 to December 2013

0.25

● Alerian Infrastructure MLPs 0.20

0.15 ● Real Estate

Returns 0.10

● High Yield

● Hedge Funds ● ● US Equities Corporate● Bonds 0.05 Government Bonds ● World Ex−US Equities ● Commodities

0.00

−0.05 0.00 0.05 0.10 0.15 0.20 0.25 0.30 Standard Deviation

Reference Assets Reference and Infrastructure Assets

Figure 6: Mean-Variance Frontier of total returns PFI portfolio and reference portfolio, 2000- 2014

0.20 Efficient Frontier January 2000 to December 2013

0.15

● PFI Portfolio 0.10

● World Ex−UK Equities Returns ● UK Equities ● UK Gilts 0.05

● Hedge Funds

● Real Estate

0.00

● Commodities −0.05 0.00 0.05 0.10 0.15 0.20 0.25 0.30 Standard Deviation

Reference Assets Reference and Infrastructure Assets

A Publication of the EDHEC Infrastructure Institute-Singapore 39 Searching for a listed infrastructure asset class- June 2016

6. Results

according to the Huberman and Kandel infrastructure creates an exposure to (1987) test, but the Kan and Zhou (2012) a combination of factors, which is not test finds that these benefits are only due otherwise available to investors already to a change in the global minimum variance allocating to the well-known factors portfolio. Without a corresponding increase described in section 5.4. in the tangency portfolio, it is not possible to conclude that the efficient frontier has As above we first present our results been improved. Still, these results may be for listed infrastructure companies considered inconclusive, as PFI portfolio (section 6.2.1), followed by global listed returns begin in 2006. infrastructure (section 6.2.2) and US (section 6.2.3) infrastructure indices. Unfor- After the GFC however, the price returns tunately, as this stage, we cannot build a of the PFI portfolio, pass the Huberman reference factor portfolio for the UK for and Kandel (1987) test, but the Kan and lack of sufficient data. Zhou (2012) finds that this is only due to the improvement of the minimum variance 6.2.1 Listed infrastructure companies portfolio but not the tangency portfolio. Table 9 presents the our results for the However, the total returns PFI portfolio is infrastructure portfolios using the naïve found to exhibit diversification benefits by infrastructure definition proposed in both the Huberman and Kandel (1987) and section 5. Using the full sample (panels Kan and Zhou (2012) tests. A and D in table 9), the Huberman and Kandel (1987) rejects the null hypothesis Hence, the impact of the PFI portfolio that the efficient frontier is not improved appears to be one of the most persistent in five of the nine price return indices and of the various ”infrastructure” portfolios 6 of the nine total return indices. Applying that were tested on a total returns basis. the Kan and Zhou (2012) test however, It improves diversification for the entire there is no evidence that infrastructure, investment period and, crucially, post-GFC, thus defined, provides diversification when all but one of the other infras- benefits. Indices that qualified under the tructure indices fails to pass the post- Huberman and Kandel (1987) test all fail GFC test of persistence. to reject the null hypothesis for both steps of the Kan and Zhou (2012) test. Consistent with the findings for the asset 6.2 Factor-based mean-variance class reference portfolio, the addition spanning test results of listed infrastructure companies to a Next, we examine how the different listed factor-based allocation does not improve infrastructure definitions proposed above the mean-variance frontier. fare against a factor-based reference portfolio, i.e. whether investing in listed

40 A Publication of the EDHEC Infrastructure Institute-Singapore Searching for a listed infrastructure asset class - June 2016

6. Results

Pre- and post-GFC results are consistent Dow Jones Brookfield and FTSE Core Infras- with the full sample. In the period January tructure index can be said to improve the 2000 to December 2008 (panels B and E reference efficient frontier. in table 9), eight of the nine price return indices are found to improve the efficient For the period January 2000 to December frontier according to the Huberman and 2008 (panels B in table 10), only four of the Kandel (1987) test. When the Kan and Zhou eight indices are found to reject the null of (2012) test is applied however, this positive the Huberman and Kandel (1987) test at the result is overturned with none of the indices 5% level, but none of these pass the Kan and examined passing the two-stage test. When Zhou (2012) test. Between January 2009 total returns are employed the results are and December 2014 (panels C in table 10) the same. only two of the eight portfolios are found to reject the null of the Huberman and Kandel For the period January 2009 to December (1987) test at the 5% level. Of these, only the 2014 (panels C and F in table 9), results Dow Jones Brookfield is found to improve mirror the pre-GFC sample. For the price the efficient frontier by the Kan and Zhou return indices the Huberman and Kandel (2012) test. (1987) test finds that the mean variance frontier is improved in six of the naïve Using total returns for the full sample period infrastructure portfolios. However, the Kan (panels D in table 10), all infrastructure and Zhou (2012) test results do not support indices examined reject the null of the these findings and none of the portfolios Huberman and Kandel (1987) test at the 5% qualify. The total returns for naïve infras- level; and four still pass the Kan and Zhou tructure portfolios result in the same (2012) test: the S&P Global Infrastructure, conclusions. FTSE Macquarie Index, MSCI ACWI Capped Index and the UBS 50-50 Index. 6.2.2 Global infrastructure indices The results for the spanning tests for global The same is true when the period January listed infrastructure indices are presented in 2000 to December 2008 (panels E in table 10. The results are now familiar. table 10), is considered: all indices pass the Huberman and Kandel (1987) test and Using price returns for the full sample three (The FTSE Core, MSCI World Infras- (panels A in table 10), six of the eight indices tructure and MSCI ACWI Infrastructure) are examined reject the null of the Huberman found by the Kan and Zhou (2012) test and Kandel (1987) test at the 5% level, to improve the tangency portfolio and the but the Kan and Zhou (2012) test indicates global minimum variance portfolio, with The that only two of the eight indices improve remainder are found to only improve the both the tangency portfolio as well as the tangency portfolio. global minimum variance portfolio: only the

A Publication of the EDHEC Infrastructure Institute-Singapore 41 Searching for a listed infrastructure asset class- June 2016

6. Results

However, from January 2009 to December 2009 to December 2014 (Panel F in Table 11), 2014, only three of the eight portfolios pass the Alerian MLP index passes both the the Huberman and Kandel (1987) test and Huberman and Kandel (1987) and Kan only one (the MSCI ACWI Capped Index) is and Zhou (2012) tests, indicating that the found to improve the efficient frontier by efficient frontier has been improved. the Kan and Zhou (2012) test on a total return basis. Hence, the MLP index is found to have a somewhat similar spanning profile than the 6.2.3 U.S. infrastructure indices PFI portfolio in the sense that it manages Finally, Table 11 shows the same results to create diversification benefits both before using U.S. market indices and factors. For the after the GFC when considered from a total full sample period and using price returns return perspective. (Panel A in Table 11), the Alerian MLP Infras- tructure index is, again, the only index found to improve the efficient frontier using the Huberman and Kandel (1987) test.

In the period from January 2000 to December 2008 (Panel B in Table 11), the Alerian MLP Infrastructure Index rejects the null hypothesis of the Huberman and Kandel (1987) test at the 5% level, but the Kan and Zhou (2012) test concludes that only the tangency portfolio has improved. In the post-GFC period (Panel C in Table 11), similar conclusions hold

Using total returns for the full sample period (Panel D in Table 11), only the Alerian MLP Infrastructure Index is again found to pass the Huberman and Kandel (1987) test, but the results of the Kan and Zhou (2012) test indicate that this is due to the Alerian MLP Infrastructure Index only improving the tangency portfolio.

From January 2000 to December 2008 (Panel E in Table 11), conclusions are the same. However, for the period from January

42 A Publication of the EDHEC Infrastructure Institute-Singapore Searching for a listed infrastructure asset class - June 2016

6. Results 3760 4229 2513 7661 6894 5407 0062 0039 1824 6945 8083 4123 4895 2844 5977 0612 0264 4281 9837 6453 3248 2671 1606 3768 5060 9767 8158 3654 0590 6754 7195 1584 2803 9174 1591 6357 ...... 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 90% Rev Util. 0 0 1 90% Rev Util. 0 0 0 90% Rev Util. 5 8 1 90% Rev Util. 0 0 0 90% Rev Util. 0 1 0 90% Rev Util. 2 5 0 8923 6639 8440 5707 9379 2896 1697 7946 0611 4272 2026 7870 3265 6029 1597 0496 5043 0179 1140 1894 0388 5641 0061 1333 8230 0683 6290 8547 6359 0732 1320 2724 0062 1475 4509 8931 ...... 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 90% Rev Transp. 0 0 0 90% Rev Transp. 0 0 1 90% Rev Transp. 1 0 3 90% Rev Transp. 0 1 0 90% Rev Transp. 1 0 2 90% Rev Transp. 3 0 5 0098 0036 3862 2662 1210 6297 0906 0393 4673 0577 0293 3307 4330 2067 7884 7544 5787 6136 7566 7251 7545 3413 4454 2339 4927 4232 5344 8997 8271 9514 8439 6151 0724 2831 3114 2574 ...... 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 90% Rev Tel. 4 8 0 90% Rev Tel. 1 2 0 90% Rev Tel. 2 4 0 90% Rev Tel. 2 4 0 90% Rev Tel. 0 1 0 90% Rev Tel. 0 0 0 5787 . 3150 5391 1639 5622 5328 3819 0015 0011 1204 4975 6926 2649 3203 2047 4152 0232 0113 3010 1629 3787 9541 5791 3918 7712 2259 4763 7009 1568 2512 1515 6295 6693 9925 8025 0868 ...... 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 75% Rev Util. 1 0 1 75% Rev Util. 0 0 0 75% Rev Util. 7 11 2 75% Rev Util. 0 0 1 75% Rev Util. 1 1 0 75% Rev Util. 3 6 1 ...... 9336 2620 5852 . . . 4460 0001 0605 8060 0000 1403 0075 1091 7419 0018 5677 2163 2542 6159 9142 0925 9398 0001 3657 5462 0000 7298 0109 0220 8824 5298 0026 0172 0558 7058 1961 2828 0424 ...... 75% Rev Transp 9 0 0 0 18 0 75% Rev Transp 5 0 0 0 10 0 75% Rev Transp 1 0 0 0 2 0 75% Rev Transp 9 0 0 0 19 0 75% Rev Transp 4 0 0 0 9 0 75% Rev Transp 3 0 1 0 4 0 0132 0048 4127 2101 1041 4944 0854 1001 1368 0613 0263 4264 2985 1606 5061 5919 9842 3040 4346 1672 6742 5847 6911 4704 5562 7828 2683 8372 0241 6356 2237 9978 4452 5286 0004 0731 ...... 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 75% Rev Tel. 4 8 0 75% Rev Tel. 1 2 0 75% Rev Tel. 2 2 2 75% Rev Tel. 2 5 0 75% Rev Tel. 1 1 0 75% Rev Tel. 0 0 1 5665 . 4685 8408 0981 3148 1516 4820 4648 5052 7730 0069 5479 6647 8781 4519 7507 7268 5531 2331 3604 1493 7307 6979 4891 0002 0005 0375 4632 9340 2151 5167 3510 5030 0119 0117 1149 ...... 50% Rev Util. 1 0 2 50% Rev Util. 0 0 0 50% Rev Util. 9 13 4 50% Rev Util. 0 0 1 50% Rev Util. 0 0 0 50% Rev Util. 4 6 2 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 4399 2398 6673 4151 2168 1383 0327 3598 5691 4524 5027 2231 1961 8224 3829 5382 0094 9229 0944 3371 0279 8677 1730 1423 6169 5416 2847 5948 9560 3305 0241 9762 0000 9982 0490 8255 ...... 50% Rev Transp. 1 0 0 0 2 0 50% Rev Transp. 1 0 0 0 1 0 50% Rev Transp. 0 0 0 0 0 0 50% Rev Transp. 1 0 0 0 2 0 50% Rev Transp. 0 0 0 0 0 0 50% Rev Transp. 0 0 0 0 0 0 Table 5: Mean-variance spanning results for Naïve infrastructure portfolios with asset class-built reference portfolio 4124 . 9995 5534 7789 7052 8225 9650 1676 7319 1227 8185 4109 3508 7987 8786 4033 0182 8004 0030 0009 4579 0669 0324 3666 1484 1458 1927 0178 0058 5223 1005 0541 3508 6698 8932 3742 ...... 50% Rev Tel. 5 11 0 50% Rev Tel. 2 4 0 50% Rev Tel. 1 2 1 50% Rev Tel. 4 7 0 50% Rev Tel. 2 3 0 50% Rev Tel. 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 Panel A Panel B Panel C Panel D Panel E Panel F , H&K Stepdown1 Stepdown2 , H&K Stepdown1 Stepdown2 , H&K Stepdown1 Stepdown2 , H&K Stepdown1 Stepdown2 , H&K Stepdown1 Stepdown2 , H&K Stepdown1 Stepdown2 p-value p-value p-value p-value p-value p-value p-value p-value p-value p-value p-value p-value p-value p-value p-value p-value p-value p-value Total returns Price returns

A Publication of the EDHEC Infrastructure Institute-Singapore 43 Searching for a listed infrastructure asset class- June 2016

6. Results 0508 2130 0356 0852 0816 1696 2699 4802 1441 0086 0385 0221 0241 0217 1413 2943 8255 1201 0324 5627 4876 5240 0950 9133 3365 5041 1854 8912 3504 3294 8680 4413 1979 2466 0490 4799 ...... 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 UBS 50-50 3 1 4 UBS 50-50 2 3 1 UBS 50-50 1 0 2 UBS 50-50 4 4 5 UBS 50-50 3 5 2 UBS 50-50 1 0 2 4546 . 8483 0005 0417 0459 0009 5502 0052 2296 0753 7009 0066 1329 1267 2978 5872 0106 0493 9542 0002 9378 0054 5882 0023 2768 0027 6679 0192 5815 0118 5662 0846 9395 3360 1967 0445 ...... UBS 7 0 4 0 11 0 UBS 5 0 3 0 7 0 UBS 2 0 0 0 4 0 UBS 8 0 7 0 9 0 UBS 6 0 5 0 6 0 UBS 2 0 0 0 4 0 4068 9118 5604 8996 6365 7925 4718 5811 ...... 6229 0287 0735 4940 6966 3858 7157 7857 5961 8247 0000 1072 0000 0001 1575 0000 0530 4847 0199 0000 0043 0000 0000 0323 0001 0278 0624 0547 ...... MSCI ACWI Capped 16 2 29 MSCI ACWI Capped 10 2 18 MSCI ACWI Capped 3 0 5 MSCI ACWI Capped 17 8 25 MSCI ACWI Capped 11 4 17 MSCI ACWI Capped 3 3 3 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1364 0459 9114 0495 3224 1293 9377 0576 7302 3948 1589 0252 8007 4534 5921 2115 0092 9240 0253 1351 8027 3715 2516 0731 9696 0559 0282 8669 9684 0163 1752 8397 0143 9051 3411 5612 ...... MSCI World Infra 3 0 3 0 2 0 MSCI World Infra 2 0 0 0 5 0 MSCI World Infra 0 0 1 0 0 0 MSCI World Infra 2 0 0 0 3 0 MSCI World Infra 2 0 0 0 5 0 MSCI World Infra 0 0 0 0 0 0 1562 2090 1011 2605 4140 0090 2095 4114 0078 8989 9974 7490 9031 3744 2478 1776 1403 2113 3188 2742 2965 1230 1315 1670 8115 5235 9300 0597 0483 1890 0715 0769 1446 3145 1483 6473 ...... FTSE Core 1 1 1 FTSE Core 2 2 2 FTSE Core 0 0 0 FTSE Core 2 3 1 FTSE Core 2 3 2 FTSE Core 1 2 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1041 0475 0125 9110 2323 0135 4154 0147 0874 2997 7363 0065 2867 2831 5164 1175 0557 8141 5856 0115 8898 3469 2868 0045 0174 0035 9175 0909 9370 0036 2773 7587 5185 4741 0364 8493 ...... FTSE Macquarie 3 0 0 0 6 0 FTSE Macquarie 4 0 1 0 7 0 FTSE Macquarie 1 0 2 0 0 0 FTSE Macquarie 4 0 0 0 8 0 FTSE Macquarie 6 0 2 0 8 0 FTSE Macquarie 0 0 0 0 0 0 4252 8950 3104 8824 4921 4168 0563 ...... 0823 6243 9489 1947 8322 5613 9554 0673 3337 0031 7051 0001 7746 0000 0004 3331 0001 3093 3650 2159 0000 0877 0000 0000 0472 0001 2706 9556 1048 ...... S&P Global Infra 10 0 20 S&P Global Infra 8 0 16 S&P Global Infra 1 0 1 S&P Global Infra 13 2 24 S&P Global Infra 11 4 18 S&P Global Infra 1 0 2 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 2379 2128 0749 0164 8074 1735 7596 7266 6421 7138 9255 0480 1011 1992 8393 9270 1569 2964 0026 0420 0052 0560 1836 0451 0285 0579 0607 0003 0056 0031 0211 0784 0314 0043 0094 0421 Table 6: Mean-variance spanning test results for global listed infrastructure indices with asset class-built reference portfolio ...... Dow Jones Brookfield 6 4 8 Dow Jones Brookfield 3 1 4 Dow Jones Brookfield 3 3 3 Dow Jones Brookfield 8 7 9 Dow Jones Brookfield 4 3 4 Dow Jones Brookfield 5 7 4 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 Panel A Panel B Panel C Panel D Panel E Panel F , H&K Stepdown1 Stepdown2 , H&K Stepdown1 Stepdown2 , H&K Stepdown1 Stepdown2 , H&K Stepdown1 Stepdown2 , H&K Stepdown1 Stepdown2 , H&K Stepdown1 Stepdown2 p-value p-value p-value p-value p-value p-value p-value p-value p-value p-value p-value p-value p-value p-value p-value p-value p-value p-value Total returns Price returns

44 A Publication of the EDHEC Infrastructure Institute-Singapore Searching for a listed infrastructure asset class - June 2016

6. Results

Table 7: Mean Variance Span test results for the U.S. listed infrastructure indices with asset class-built reference portfolio

Price returns , Panel A Alerian MLP FTSE Macquarie USA MSCI USA Infrastructure MSCI USA Infra. Capped H&K 3.5537 0.9730 0.4046 1.7772 p-value 0.0308 0.3821 0.6679 0.1722 Stepdown1 6.8957 0.0604 0.0512 0.3592 p-value 0.0094 0.8064 0.8212 0.5498 Stepdown2 0.2048 1.9064 0.7622 3.2072 p-value 0.6515 0.1709 0.3839 0.0751 , Panel B Alerian MLP FTSE Macquarie USA MSCI USA Infrastructure MSCI USA Infra. Capped H&K 1.9041 0.3161 0.4258 2.2540 p-value 0.1544 0.7349 0.6545 0.1104 Stepdown1 3.7423 0.5859 0.0962 0.1576 p-value 0.0559 0.4588 0.7570 0.6922 Stepdown2 0.0640 0.0477 0.7623 4.3878 p-value 0.8008 0.8305 0.3847 0.0388 , Panel C Alerian MLP FTSE Macquarie USA MSCI USA Infrastructure MSCI USA Infra. Capped H&K 0.8072 0.1190 0.0891 0.0867 p-value 0.4507 0.8880 0.9148 0.9171 Stepdown1 1.6137 0.0577 0.1382 0.0350 p-value 0.2086 0.8109 0.7114 0.8523 Stepdown2 0.0007 0.1830 0.0407 0.1405 p-value 0.9797 0.6702 0.8408 0.7090 Total returns , Panel D Alerian MLP FTSE Macquarie USA MSCI USA Infrastructure MSCI USA Infra. Capped H&K 6.0024 1.1463 0.6895 1.9286 p-value 0.0030 0.3227 0.5032 0.1485 Stepdown1 11.6155 0.0121 0.0246 0.9511 p-value 0.0008 0.9128 0.8756 0.3308 Stepdown2 0.3666 2.3070 1.3622 2.9069 p-value 0.5456 0.1325 0.2448 0.0900 , Panel E Alerian MLP FTSE Macquarie USA MSCI USA Infrastructure MSCI USA Infra. Capped H&K 3.8572 0.2963 0.5864 2.3219 p-value 0.0244 0.7488 0.5583 0.1035 Stepdown1 7.4502 0.4998 0.0063 0.2440 p-value 0.0075 0.4931 0.9368 0.6225 Stepdown2 0.2480 0.0966 1.1783 4.4336 p-value 0.6196 0.7609 0.2803 0.0378 , Panel F Alerian MLP FTSE Macquarie USA MSCI USA Infrastructure MSCI USA Infra. Capped H&K 0.7444 0.1307 0.4085 0.3702 p-value 0.4791 0.8777 0.6664 0.6921 Stepdown1 1.3711 0.0963 0.7868 0.3305 p-value 0.2460 0.7574 0.3784 0.5674 Stepdown2 0.1171 0.1674 0.0304 0.4143 p-value 0.7333 0.6838 0.8622 0.5221

Table 8: Mean-variance spanning test results for the PFI stocks with asset class-built reference portfolio

Price returns Total returns , Full Sample , Full Sample H&K 0.4211 H&K 5.7016 p-value 0.6575 p-value 0.0045 Stepdown1 0.0000 Stepdown1 5.4260 p-value 0.9997 p-value 0.0219 Stepdown2 0.8508 Stepdown2 5.7239 p-value 0.3586 p-value 0.0186 , Pre-GFC , Pre-GFC H&K 0.4173 H&K 1.3958 p-value 0.6630 p-value 0.2649 Stepdown1 0.6413 Stepdown1 1.4184 p-value 0.4302 p-value 0.2440 Stepdown2 0.1959 Stepdown2 1.3529 p-value 0.6615 p-value 0.2546 , Post-GFC , Post-GFC H&K 7.9863 H&K 7.6910 p-value 0.0008 p-value 0.0010 Stepdown1 1.1721 Stepdown1 9.1745 p-value 0.2830 p-value 0.0035 Stepdown2 14.7621 Stepdown2 5.5234 p-value 0.0003 p-value 0.0218

A Publication of the EDHEC Infrastructure Institute-Singapore 45 Searching for a listed infrastructure asset class- June 2016

6. Results 3801 8147 1698 1972 4114 1084 9726 0551 9003 6385 6781 6038 0764 3294 0402 4792 2809 5826 0406 1867 0306 6859 5053 5781 6374 9605 3160 7437 1817 3049 3058 7675 8083 3791 4485 3123 ...... 0 0 0 0 0 0 90% Rev Util. 0 0 1 90% Rev Util. 1 0 2 0 0 0 0 0 0 0 0 0 0 0 0 90% Rev Util. 2 0 4 90% Rev Util. 0 1 0 90% Rev Util. 3 1 4 90% Rev Util. 0 0 0 0045 8754 . . 0383 1746 0303 0102 0464 0225 3250 8583 7685 7065 0237 2982 0019 5427 0005 0286 2631 0152 0016 3416 0005 0078 0690 0112 6493 3731 7488 2738 1988 8887 9129 2298 4145 8036 ...... 0 0 0 0 0 0 90% Rev Transp. 3 1 4 90% Rev Transp. 4 4 5 0 0 0 0 0 0 0 0 0 0 0 0 90% Rev Transp. 6 0 13 90% Rev Transp. 3 1 6 90% Rev Transp. 6 0 12 90% Rev Transp. 5 3 6 0500 4815 3526 3942 6323 2949 7121 4482 4179 . . . . 0000 8644 0000 0000 4720 0000 0293 5196 . . . . . 0000 3254 0000 0021 8542 0004 0000 1496 0000 0293 5496 0094 9763 7757 0340 1082 7247 3616 1550 ...... 0 0 0 0 0 0 90% Rev Tel. 37 0 74 90% Rev Tel. 38 0 76 0 0 0 0 0 0 0 0 0 0 0 0 90% Rev Tel. 14 0 28 90% Rev Tel. 6 0 13 90% Rev Tel. 18 2 34 90% Rev Tel. 3 0 7 2523 7382 1036 1276 3654 0690 3880 1121 6775 0835 8237 3467 0480 2324 0308 4150 1965 7794 0262 1281 0256 6593 3850 7866 1284 4435 7927 8911 7022 0791 7743 3533 1278 4192 7647 0739 ...... 0 0 0 0 0 0 75% Rev Util. 1 0 2 75% Rev Util. 2 0 3 0 0 0 0 0 0 0 0 0 0 0 0 75% Rev Util. 3 1 4 75% Rev Util. 0 1 0 75% Rev Util. 3 2 5 75% Rev Util. 0 0 0 ...... 0324 0150 3365 0056 0023 3064 4973 0392 9289 3390 5743 0524 5564 2159 5529 2156 5515 2157 2293 0458 3858 0702 9687 0894 0385 1355 5522 1132 5023 2231 5354 0142 7151 0196 1381 0810 ...... 0 0 0 0 0 0 75% Rev Transp 3 6 0 75% Rev Transp 5 9 1 75% Rev Transp 1 0 1 0 1 0 75% Rev Transp 3 0 3 0 2 0 75% Rev Transp 2 0 2 0 1 0 75% Rev Transp 4 0 5 0 3 0 8602 1880 8485 1603 7341 2956 8691 3121 9038 . . . . 0000 9323 0000 0000 4064 0000 0072 6925 . . . . . 0000 2912 0000 0060 5800 0016 0000 1405 0000 0259 5131 0085 1258 5339 3092 2061 8623 4323 3537 ...... 0 0 0 0 0 0 75% Rev Tel. 41 0 84 75% Rev Tel. 44 0 89 0 0 0 0 0 0 0 0 0 0 0 0 75% Rev Tel. 19 1 38 75% Rev Tel. 5 0 10 75% Rev Tel. 23 2 43 75% Rev Tel. 3 0 7 1545 7781 0555 0870 5745 0323 8876 0796 7149 4766 3163 6550 ...... 5904 8145 3084 8455 7212 7121 2056 0863 2379 3435 2420 3352 0123 1809 0080 0108 0196 0582 0070 1517 0050 0413 0763 0722 ...... 0 0 0 0 0 0 50% Rev Util. 1 0 3 50% Rev Util. 2 0 4 50% Rev Util. 4 1 7 50% Rev Util. 4 5 3 50% Rev Util. 5 2 8 50% Rev Util. 3 3 3 0 0 0 0 0 0 0 0 0 0 0 0 0747 2978 0425 0174 2298 0098 6335 0904 1744 1465 4520 8235 ...... 4947 0340 1873 6661 8562 0102 0454 1373 2274 6350 9078 0521 4858 0136 1478 7014 8974 0036 6536 1991 9210 3407 3890 1268 ...... 0 0 0 0 0 0 50% Rev Transp. 2 1 4 50% Rev Transp. 4 1 6 50% Rev Transp. 3 0 0 0 6 0 50% Rev Transp. 2 0 0 0 3 0 50% Rev Transp. 4 0 0 0 8 0 50% Rev Transp. 1 0 0 0 2 0 8309 1669 0714 6941 . . . . 1594 8786 4170 2175 8549 0000 8702 0000 0000 4228 0000 0268 6454 ...... 3241 5824 0032 8219 2249 6878 8071 0000 2526 0000 0005 9548 0001 0000 0960 0000 0078 4099 0026 ...... 0 0 0 0 0 0 50% Rev Tel. 47 0 96 50% Rev Tel. 49 0 97 50% Rev Tel. 19 1 36 50% Rev Tel. 8 0 17 50% Rev Tel. 23 2 42 50% Rev Tel. 5 0 9 0 0 0 0 0 0 0 0 0 0 0 0 Table 9: Mean Variance Span test results for the factor asset class and the Naïve infrastructure portfolios with factor-based reference portfolios Full Sample Pre-GFC Post-GFC Full Sample Pre-GFC Post-GFC p-value p-value p-value p-value p-value p-value , H&K Stepdown1 Stepdown2 , H&K Stepdown1 Stepdown2 , H&K Stepdown1 Stepdown2 , H&K Stepdown1 Stepdown2 , H&K Stepdown1 Stepdown2 , H&K Stepdown1 Stepdown2 p-value p-value p-value p-value p-value p-value p-value p-value p-value p-value p-value p-value Total returns Price returns

46 A Publication of the EDHEC Infrastructure Institute-Singapore Searching for a listed infrastructure asset class - June 2016

6. Results 4645 9591 1227 . . . 0324 0508 5627 2130 4876 0356 0001 0011 4302 0025 1776 0666 7944 7735 4760 0684 0852 0816 1696 0004 0003 1370 1043 0400 6015 5240 0950 9133 3279 2468 3387 3885 2753 ...... UBS 50-50 3 0 1 0 4 0 UBS 50-50 10 0 10 0 9 0 0 0 0 UBS 50-50 1 3 0 0 0 0 0 0 0 0 0 0 UBS 50-50 2 3 1 UBS 50-50 8 14 2 UBS 50-50 2 4 0 4546 . 8483 0005 0417 0459 0009 9037 0220 7860 0059 0207 8857 1224 0415 9025 1674 3189 0151 5502 0052 2296 0753 7009 0066 0881 0196 0081 0056 1573 6925 5165 0353 9994 0102 0309 8610 ...... UBS 7 0 4 0 11 0 UBS 3 0 7 0 0 0 0 0 0 UBS 2 4 0 UBS 5 0 3 0 7 0 UBS 4 0 8 0 0 0 UBS 3 0 6 0 0 0 4068 9118 5604 8996 . . 0000 6229 1072 0000 6688 0001 5355 0023 3438 0026 0155 1444 0125 4366 1816 5773 . . 0287 4560 8566 6552 9643 9264 6012 0001 1575 0000 0010 0022 0333 0041 0105 0355 ...... MSCI ACWI Capped 16 0 2 0 29 0 MSCI ACWI Capped 9 0 9 0 9 0 0 0 0 MSCI ACWI Capped 4 2 6 MSCI ACWI Capped 10 2 18 MSCI ACWI Capped 7 9 4 MSCI ACWI Capped 5 6 4 0 0 0 0 0 0 0 0 0 4659 9290 4088 0790 . . 0459 0495 1293 0000 1714 0000 1364 9114 3224 8860 0657 4790 0257 8372 5068 2053 . . 9377 0576 7302 3948 1589 0252 0000 4843 0366 0000 0381 0546 3362 1311 6679 0597 ...... 0 0 0 0 0 0 MSCI World Infra 3 3 2 MSCI World Infra 12 1 22 0 0 0 MSCI World Infra 2 0 5 MSCI World Infra 2 0 0 0 5 0 MSCI World Infra 21 0 4 0 37 0 MSCI World Infra 3 0 2 0 3 0 3188 2742 2965 0073 0075 1036 1562 2090 1011 1611 4542 6972 1620 1262 2553 8708 3981 3164 2605 4140 0090 2948 1231 5010 6625 2455 0304 1230 1315 1670 0007 0051 0101 0771 0432 3137 ...... 0 0 0 0 0 0 FTSE Core 1 1 1 FTSE Core 5 7 2 0 0 0 FTSE Core 1 2 1 FTSE Core 2 2 2 FTSE Core 9 9 7 FTSE Core 2 4 1 0 0 0 0 0 0 0 0 0 0475 9110 0135 0039 0144 0238 1041 0125 2323 7400 1167 2047 9662 8081 9224 0344 0595 0096 ...... 4154 0147 0874 2997 7363 0065 5794 0126 7027 0040 4225 5172 3480 7074 6576 4203 0386 8448 ...... 0 0 0 0 0 0 FTSE Macquarie 3 0 6 FTSE Macquarie 5 6 5 0 0 0 FTSE Macquarie 0 0 0 FTSE Macquarie 4 0 1 0 7 0 FTSE Macquarie 4 0 8 0 0 0 FTSE Macquarie 0 0 0 0 0 0 4252 8950 1028 6201 . . . . 3104 6604 0001 7746 0000 0000 0001 0058 0823 8277 5485 3215 6447 6060 9974 2146 ...... 6243 9489 3514 0354 2262 0731 3734 0004 3331 0001 0005 0001 8512 2999 1546 5432 ...... 0 0 0 0 0 0 S&P Global Infra 10 0 20 S&P Global Infra 12 15 7 0 0 0 S&P Global Infra 0 0 0 S&P Global Infra 8 0 16 S&P Global Infra 8 16 0 S&P Global Infra 1 2 0 0 0 0 0 0 0 0 0 0 0756 . 2588 0624 0026 0420 0052 0002 0000 4559 2379 2128 0749 3519 5590 0026 0051 0443 5136 3745 1975 Table 10: Mean Variance Span test results for the Global listed infrastructure indices with factor-based reference portfolios ...... 0164 8074 1735 0486 7087 3691 2020 0560 1836 0451 0017 0005 4029 0013 0014 0780 ...... 0 0 0 0 0 0 Dow Jones Brookfield 6 4 8 Dow Jones Brookfield 9 18 0 0 0 0 Dow Jones Brookfield 6 8 4 Dow Jones Brookfield 3 1 4 Dow Jones Brookfield 7 13 0 Dow Jones Brookfield 7 11 3 0 0 0 0 0 0 0 0 0 Full Sample Pre-GFC Post-GFC A Full Sample Pre-GFC Post-GFC p-value p-value p-value p-value p-value p-value p-value p-value p-value , H&K Stepdown1 Stepdown2 , H&K Stepdown1 Stepdown2 , H&K Stepdown1 Stepdown2 , H&K Stepdown1 Stepdown2 , H&K Stepdown1 Stepdown2 , H&K Stepdown1 Stepdown2 p-value p-value p-value p-value p-value p-value p-value p-value p-value Total returns Price returns

A Publication of the EDHEC Infrastructure Institute-Singapore 47 Searching for a listed infrastructure asset class- June 2016

6. Results

Table 11: Mean Variance Span test results for the factor asset classes and U.S. listed infrastructure indices with factor-based reference portfolios

Price returns , Full Sample Alerian MLP FTSE Macquarie USA MSCI USA Infrastructure MSCI USA Infra. Capped H&K 8.0362 1.0298 1.0553 0.1836 p-value 0.0005 0.3613 0.3503 0.8324 Stepdown1 8.7670 0.0449 0.5795 0.2557 p-value 0.0035 0.8326 0.4475 0.6137 Stepdown2 6.9949 2.0365 1.5348 0.1120 p-value 0.0089 0.1571 0.2170 0.7383 , Pre-GFC Alerian MLP FTSE Macquarie USA MSCI USA Infrastructure MSCI USA Infra. Capped H&K 3.2144 1.3422 0.0637 0.6650 p-value 0.0443 0.2909 0.9383 0.5165 Stepdown1 6.4173 0.0560 0.1059 0.6719 p-value 0.0128 0.8161 0.7455 0.4143 Stepdown2 0.0109 2.7932 0.0216 0.6601 p-value 0.9171 0.1141 0.8833 0.4184 , Post-GFC Alerian MLP FTSE Macquarie USA MSCI USA Infrastructure MSCI USA Infra. Capped H&K 0.8072 0.1190 0.0891 0.0867 p-value 0.4507 0.8880 0.9148 0.9171 Stepdown1 1.6137 0.0577 0.1382 0.0350 p-value 0.2086 0.8109 0.7114 0.8523 Stepdown2 0.0007 0.1830 0.0407 0.1405 p-value 0.9797 0.6702 0.8408 0.7090 Total returns , Full Sample Alerian MLP FTSE Macquarie USA MSCI USA Infrastructure MSCI USA Infra. Capped H&K 12.4051 0.9491 1.3303 0.1821 p-value 0.0000 0.3910 0.2671 0.8337 Stepdown1 21.9107 0.1018 0.3802 0.3176 p-value 0.0000 0.7505 0.5383 0.5738 Stepdown2 2.5900 1.8147 2.2886 0.0468 p-value 0.1093 0.1814 0.1321 0.8289 , Pre-GFC Alerian MLP FTSE Macquarie USA MSCI USA Infrastructure MSCI USA Infra. Capped H&K 5.4190 0.9820 0.0060 0.5267 p-value 0.0058 0.3974 0.9940 0.5921 Stepdown1 9.8783 0.0571 0.0001 0.0966 p-value 0.0022 0.8144 0.9938 0.7565 Stepdown2 0.8830 2.0263 0.0121 0.9654 p-value 0.3496 0.1738 0.9127 0.3282 , Post-GFC Alerian MLP FTSE Macquarie USA MSCI USA Infrastructure MSCI USA Infra. Capped H&K 16.3809 0.8787 1.0270 1.6919 p-value 0.0000 0.4201 0.3637 0.1921 Stepdown1 8.4217 1.2882 0.8764 0.9380 p-value 0.0050 0.2605 0.3526 0.3363 Stepdown2 21.9127 0.4672 1.1798 2.4481 p-value 0.0000 0.4966 0.2813 0.1224

48 A Publication of the EDHEC Infrastructure Institute-Singapore 7. Conclusions

A Publication of the EDHEC Infrastructure Institute-Singapore 49 Searching for a listed infrastructure asset class- June 2016

7. Conclusions

7.1 Summary of results in investors’ portfolios. Expecting the In this paper, we examined the contention emergence of a new or unique ”infras- that focusing on ”listed infrastructure” tructure asset class” by focusing on public has the potential to create diversification equities selected on the basis of industrial benefits previously unavailable to large sectors is thus misguided. Such asset investors already active in public markets. selection schemes do not create diver- The reasons for doing so were threefold: sification benefits. Figure 4 provides an illustration of these results in the case of 1. Several papers have argued that it is the the FTSE Macquarie Listed Infrastructure case but do not provide robust statistical Index for the U.S. market. tests of this hypothesis; 2. Index providers have created dedicated Our test result are summarised in table 12. indices focusing on this theme and a Stylised facts include: number of active managers propose to invest in listed infrastructure arguing 1. We tested 22 proxies of listed infras- that it is an asset class in its own right; tructure and found little to no robust 3. Capital market instruments are often evidence of a ”listed infrastructure asset used by investors to proxy investments class” that was not already spanned by in privately-held (unlisted) infrastructure a combination of capital market instru- but the adequacy of such proxies remains ments and alternatives or a factor-based untested. asset allocation; 2. The majority of test portfolios that We tested the notion that there is a unique improved the mean-variance efficient and persistent ”listed infrastructure effect” frontier before the GFC fail to repeat this using 22 listed infrastructure proxies and a feat post GFC. There is no evidence of series of statistical tests of mean-variance persistent diversification benefits; spanning against reference portfolios, built 3. Of the 22 test portfolios used in this with either traditional asset classes or paper to try and establish the existence investment factors. We conducted these of a listed infrastructure asset class, only tests for global, U.S. and UK markets four manage to improve on a typical covering the past 15 years, on a price return asset allocation defined either by tradi- and total return basis. tional asset class or by factor exposure after the GFC and only one is not spanned We conclude that listed infrastructure, both pre- and post-GFC. We return to as it is traditionally defined by SIC code these in the discussion below; and industrial sector, is not an asset class 4. Building baskets of stocks on the basis or a unique combination of market of their SIC code and proportion of factors, but instead cannot be persistently infrastructure income fails generate a distinguished from existing exposures convincing exposure to a new asset class.

50 A Publication of the EDHEC Infrastructure Institute-Singapore Searching for a listed infrastructure asset class - June 2016

7. Conclusions

Table 12: Summary table of mean-variance spanning tests: This table summarises the findings of the mean-variance spanning tests for the infrastructure proxies and different asset allocation strategies employed in this paper. ✓indicated that the infrastructure proxy passed all three spanning tests at the 5% confidence level either with reference toan asset class-build portfolio or a factor-built portfolio. The first column reports results for the whole sample from 2000 to 2014, the next two columns report pre- and post-GFC results and the fourth column highlights the proxies that show post-GFC persistence of pre-GFC improvement of the efficient portfolio frontier.

Full sample Pre-GFC Post-GFC Price Returns 50% Rev. Req. Telecom. × × × 50% Rev. Req. Transport × × × 50% Rev. Req. Utilities × × × 75% Rev. Req. Telecom. × × × 75% Rev. Req. Transport × × × 75% Rev. Req. Utilities × × × 90% Rev. Req. Telecom. × × × 90% Rev. Req. Transport × × × 90% Rev. Req. Utilities × × × Alerian MLP ✓ × × FTSE Macquarie USA × × × MSCI USA × × × MSCI USA Infra Capped × × × DJ Brookfield Global ✓ × ✓ S&P Infrastructure × ✓ × FTSE Macquarie Infra × × × FTSE Global Core × × × MSCI World Infra × ✓ × MSCI ACWI Infra Capped × × × UBS Global Infra Uti ✓ ✓ × UBS Global 50-50 × × × PFI Portfolio × × × Total Returns 50% Rev. Req. Telecom. × × × 50% Rev. Req. Transport × × × 50% Rev. Req. Utilities × × × 75% Rev. Req. Telecom. × × × 75% Rev. Req. Transport × × × 75% Rev. Req. Utilities × × × 90% Rev. Req. Telecom. × × × 90% Rev. Req. Transport ✓ × × 90% Rev. Req. Utilities × × × Alerian MLP × × ✓ FTSE Macquarie USA × × × MSCI USA × × × MSCI USA Infra Capped × × × DJ Brookfield Global ✓ × ✓ FTSE Global Core × ✓ × S&P Infrastructure ✓ ✓ × FTSE Macquarie Infra ✓ ✓ × MSCI World Infra × ✓ × MSCI ACWI Infra Capped ✓ ✓ ✓ UBS Global Infra Uti ✓ ✓ × UBS Global 50-50 ✓ × × PFI Portfolio ✓ × ✓

A Publication of the EDHEC Infrastructure Institute-Singapore 51 Searching for a listed infrastructure asset class- June 2016

7. Conclusions

A more promising avenue is to focus 1. The Brookfield Dow Jones Infrastructure on underlying contractual or gover- Index: on close examination, this index nance structures that tend to maximise made a significant shift towards the dividend payout and pay dividends with oil and gas sector after the GFC and great regularity, such as the PFI or MLP benefited from the significant rise ofoil models; prices in the subsequent period. We note, 5. More generally, benchmarking unlisted without further investigation, that since infrastructure investments with thematic 2014 and the collapse of global oil prices, (industry-based) stock indices is unlikely it has experienced lacklustre perfor- to be very helpful from a pure asset mance. Hence, rather than an ”infras- allocation perspective i.e. the latter do tructure effect”, this proxy may have not exhibit a risk/return trade-off or been capturing a kind of ”oil play”; betas that large investors did not have 2. The MSCI ACWI Infrastructure Capped: access to already. this proxy is the only one which passes the spanning tests both pre- and post- GFC. It is one of the few listed infras- 7.2 Discussion tructure indices which is not simply While we conclude from testing the impact weighted by market capitalisation but is of 22 proxies, that there is no convincing instead constrained to have a maximum evidence of a listed infrastructure asset of one third of its assets invested in class, it is worthwhile examining the four Telecoms, one third in Utilities and proxies that manage to improve on the another third in energy and trans- proposed reference asset allocation after portation. Hence, it uses a very ad the GFC. hoc weighing scheme, vaguely resem- bling equal weighting, which never- Indeed, high pre-GFC Sharpe ratios that do theless improves on the market cap- not survive the 2008 credit crunch and lose weighted point of reference. Again, all statistical significance in mean-variance rather than an effect driven by a spanning tests post-GFC do not make good hypothetical ”infrastructure asset class”, candidates for an asset class or bundle it seems reasonable to assume that of factors. However, proxies that pass the portfolio weights explain the impact of 10 10 - In future research, similar test mean-variance tests since 2008 may at least this proxy; of mean-variance spanning against open the possibility of a more persistent 3. The Alerian MLP Index: this proxy and efficient or ”smart” reference indices is necessary to control for such effect. the next one only improve the reference effects. allocation post GFC in a total return The four proxies that are not already basis. Here the role played by dividend spanned by our reference portfolios in the payouts, their size and regularity relative post GFC period questions are: to other stocks are likely candidates to explain why they succeed in passing the

52 A Publication of the EDHEC Infrastructure Institute-Singapore Searching for a listed infrastructure asset class - June 2016

7. Conclusions

spanning tests. However, this index also unlisted infrastructure fund universe has proves to be high risk and correlated with similar characteristics. the energy price cycle 4. The PFI Portfolio, because it corresponds We conclude that as an asset selection to self-contained investment vehicles scheme, the notion of investing in ”infras- that receive a steady income stream tructure” (listed or not) should be under- from the public sector and have risky stood as a heuristic i.e. a mental short- but predictable operating and financing cut meant to create an exposure to certain costs, and are, by design, likely to have factors, but neither a thing nor an end in very regular dividend payouts. itself.

This last point is important since the A clear distinction can be made between observed improvement of the efficient infrastructure as a matter of public policy, frontier by adding assets such as MLPs or in which case the focus is rightly on indus- PFIs also corresponds to the beginning of trial functions, and the point of view of the very low interest rate policies introduced financial investors, who may be exposed by U.S. and U.K. central banks after the GFC. to completely different risks through In such an environment, such high-coupon investments in firms providing exactly paying assets start to exhibit previously the same industrial functions. Notional unremarkable characteristics that, mechan- grouping of assets by industrial sectors ically increases their ability to have an (transport, energy, water, etc) create very impact on the reference portfolio. little information or predictive power.

Crucially, what determines this ability to Focusing on definitions of infrastructure deliver regular and high dividend payouts investment that match the tangible or is the contractual and governance structure industrial characteristics of certain firms of the underlying businesses, not their or assets is unhelpful because it does not belonging to a given industrial sector, which take in to account the mechanisms that does not suggest any particular a priori create the potentially desirable character- dividend paying behaviour. istics of infrastructure investment. Infras- tructure investment should be construed However, it must be noted that the relatively solely as a way to buy claims on future low aggregate market capitalisation of cash flows created by specific underlying listed entities offering a ”clean” exposure to business models, themselves the product infrastructure ”business models” as opposed of long-term contractual arrangements to infrastructure industrial sectors may limit between public and private parties (or alter- the ability of investors to enjoy these natively between two private parties). potential benefits unless the far larger

A Publication of the EDHEC Infrastructure Institute-Singapore 53 8. Appendices

54 A Publication of the EDHEC Infrastructure Institute-Singapore Searching for a listed infrastructure asset class - June 2016

8. Appendices

Table 13: Descriptive statistics of annualised price and total returns of reference asset classes, 2000-2014

Panel A: Descriptive statistics of annualised price and total returns of the global reference asset classes, 2000-2014

Price Returns , Bonds Real Estate Commo Hedge Funds OECD Stocks EM Stocks Price return 0.0545 0.0152 0.0224 0.0615 −0.0010 0.0162 Risk 0.0584 0.1998 0.2355 0.0568 0.1611 0.2339 SR 0.8445 0.0506 0.0734 0.9900 −0.0371 0.0478 Total Returns , Bonds Real Estate Commo Hedge Funds OECD Stocks EM Stocks Tot. return 0.0545 0.0543 −0.0193 0.0634 0.0227 0.0436 Risk 0.0584 0.1993 0.2389 0.0567 0.1611 0.2336 SR 0.8446 0.2462 −0.1014 1.0253 0.1093 0.1645 Panel B: Descriptive statistics of annualised price and total returns of U.S. reference assets classes, 2000-2014

Price Returns , Gov Bonds Corp. Bonds High Yld Real Estate Commo. H. Funds U.S. Stocks World ex-U.S. Price return 0.0132 0.0109 −0.0062 0.0634 0.0104 0.0645 0.0293 0.0038 Risk 0.0412 0.0344 0.0984 0.0567 0.2340 0.2186 0.1565 0.1739 SR 0.1973 0.1701 −0.1130 1.0255 0.0227 0.2710 0.1546 −0.0072 Total Returns , Gov Bonds Corp. Bonds High Yld Real Estate Commo. H.Funds U.S. Stocks World ex-U.S. Tot. return 0.0545 0.0578 0.0806 0.0653 0.0104 0.1192 0.0482 0.0325 Risk 0.0419 0.0350 0.0996 0.0566 0.2340 0.2195 0.1566 0.1743 SR 1.1778 1.4997 0.7554 1.0597 0.0227 0.5179 0.2743 0.1568 Panel C: Descriptive statistics of annualised price and total returns of the U.K. reference asset classes, 2000-2014

Price Returns , Fixed Interest Real Estate Commo. H. Funds UK. Stocks World ex-UK Price return 0.0590 0.0305 0.0487 0.0126 −0.0036 0.0201 Risk 0.0500 0.2074 0.1206 0.2171 0.1420 0.1548 SR 1.0755 0.1224 0.3605 0.0347 −0.0603 0.0971 Total Returns , Fixed Interest Real Estate Commo. H. Funds UK. Stocks World ex-UK Tot. return 0.0590 0.0657 0.0506 0.0126 0.0309 0.0429 Risk 0.0498 0.2084 0.1206 0.2171 0.1423 0.1553 SR 1.0778 0.2898 0.3759 0.0347 0.1809 0.2427

Table 14: Descriptive statistics of annualised price and total returns of the reference factors, 2000-2014

Panel A: Descriptive statistics of annualised price and total returns of the global reference factors, 2000-2014

Price Returns , Eur Market U.S. Market Size Value Price return −0.0344 −0.0105 0.0522 0.0250 Risk 0.1963 0.1562 0.0713 0.0949 SR −0.1997 −0.0989 0.6587 0.2099

Total Returns , Eur Market U.S. Market Size Value Term Default Tot. return −0.0031 0.0084 0.0467 0.0421 0.0387 −0.0069 Risk 0.1962 0.1561 0.0717 0.0946 0.0758 0.2274 SR −0.0413 0.0219 0.5786 0.3904 0.4422 −0.0522 Panel B: Descriptive statistics of annualised price and total returns of the U.S. reference factors, 2000-2014

Price Returns , Market Size Value Term Default Price return 0.0162 0.0271 0.0290 −0.0022 −0.0242 Risk 0.1587 0.1058 0.1143 0.0870 0.0680 SR 0.0698 0.2081 0.2089 −0.0819 −0.4284 Total Returns , Market Size Value Term Default Tot. return 0.0348 0.0216 0.0430 0.0551 0.0330 Risk 0.1586 0.1061 0.1144 0.0871 0.0679 SR 0.1867 0.1558 0.3305 0.5723 0.4109

A Publication of the EDHEC Infrastructure Institute-Singapore 55 Searching for a listed infrastructure asset class- June 2016

8. Appendices

Table 15: List of project finance SPVs in the listed PFI portfolio

Project name Sector Investor Country Revenue Source A249 Road Roads HICL UK Unitary Charge A92 Road Roads HICL UK Unitary Charge Abbotsford Hospital Hospitals JLIF Canada Unitary Charge Abingdon -Thames Valley Gov Services INPP UK Unitary Charge Police Addiewell Prison Gov Services HICL UK Unitary Charge Alberta Schools Gov Services INPP Canada Unitary Charge Allenby and Connaught PFI Gov Services HICL UK Unitary Charge Project, UK Angel Trains Hospitals INPP UK Unitary Charge Aquasure Victorian Desali- Gov Services HICL Australia Purchase agreement nation Project, Australia Avon and Somerset Courts Gov Services JLIF UK Unitary Charge Barking and Dagenham PFI education INPP UK Unitary Charge SPV 1 Barking and Havering Clinics Hospitals BBGI UK Unitary Charge Barking and Dagenham education HICL UK Unitary Charge Schools Barnet and Haringey Clinics Hospitals BBGI UK Unitary Charge Barnet Hospital, UK Hospitals HICL UK Unitary Charge Barnet Street Lighting Gov Services JLIF UK Unitary Charge Barnsley PFI SPV 1 education INPP, JLIF UK Unitary Charge Barnsley PFI SPV 2 education INPP UK Unitary Charge Barnsley PFI SPV 3 education INPP UK Unitary Charge BBG Lakeside Hospitals INPP UK Unitary Charge Bedford Schools education BBGI UK Unitary Charge BeNEX Rail Link INPP Germany unknown Bentilee Hub Community Gov Services JLIF UK Unitary Charge Centre Bexley Schools education JLIF UK Unitary Charge Bexley, Bromley, Greenwich 1 Hospitals INPP UK Unitary Charge Bexley, Bromley, Greenwich 2 Hospitals INPP UK Unitary Charge BHH Mt Vernon Hospitals INPP UK Unitary Charge BHH Sudhury Hospitals INPP UK Unitary Charge Birmingham and Solihull LIFT Hospitals HICL UK Unitary Charge Birmingham Hospitals Hospitals HICL UK Unitary Charge Birmingham PFI SPV 1 education INPP UK Unitary Charge Bishop Auckland Hospital, UK Hospitals HICL UK Unitary Charge Bistol PFI SPV 1 education INPP UK Unitary Charge Blackburn PFI SPV 1 education INPP, HICL UK Unitary Charge Blackburn PFI SPV 2 education INPP UK Unitary Charge Blackpool Primary Care Facility Hospitals HICL UK Unitary Charge BMBF education INPP Germany Unitary Charge Boldon School education HICL UK Unitary Charge Bradford BSF Phase 2 education HICL,INPP UK Unitary Charge Bradford PFI SPV 1 education INPP UK Unitary Charge Brent, Harrow, Hillingdon Hospitals INPP UK Unitary Charge Brentwood Community Hospitals HICL UK Unitary Charge Hospital Brescia Hospital Hospitals INPP Portugal Unitary Charge Brighton Hospital, UK Hospitals HICL UK Unitary Charge Bristol BSF education JLIF UK Unitary Charge Bristol Fishponds and Hospitals INPP UK Unitary Charge Hampton House Bristol Shirehampton and Hospitals INPP UK Unitary Charge Whitchurch Brockley Social Housing PFI Gov Services JLIF UK Unitary Charge Burg Prison Gov Services BBGI Germany Unitary Charge Calderdale education INPP UK Unitary Charge Cambridgeshire PFI SPV 1 education INPP UK Unitary Charge Camden Housing Gov Services JLIF UK Unitary Charge Canning Town Social Housing Gov Services JLIF UK Unitary Charge PFI Central Middlesex Hospital, UK Hospitals HICL UK Unitary Charge Clackmannanshire Schools education BBGI UK Unitary Charge Cleveland Police Station and Gov Services JLIF UK Unitary Charge HQ Connect PFI Roads HICL UK Unitary Charge Conwy Schools, UK education HICL UK Unitary Charge Cork School of Music education HICL Ireland Unitary Charge Coventry Schools education BBGI UK Unitary Charge Croydon Schools education HICL UK Unitary Charge Darlington Schools, UK education HICL UK Unitary Charge Defence Sixth Form College, education HICL UK Unitary Charge UK Derby City PFI SPV 1 education INPP UK Unitary Charge Derby Courts Gov Services INPP UK Unitary Charge Derby Schools education INPP, HICL UK Unitary Charge Derby Schools 2 education INPP UK Unitary Charge Derbyshire PFI SPV 1 education INPP UK Unitary Charge Diabolo (T2 and T3 and T5) Rail Link INPP UK unknown

source: annual reports

56 A Publication of the EDHEC Infrastructure Institute-Singapore Searching for a listed infrastructure asset class - June 2016

8. Appendices

Table 16: List of project finance SPVs in the listed PFI portfolio (continued)

Project name Sector Investor Country Revenue Source Doncaster Mental Health Hospitals HICL UK Unitary Charge Doncaster Schools education HICL UK Unitary Charge Dorset Fire and Rescue Gov Services HICL UK Unitary Charge Dublin Courts Gov Services INPP Ireland Unitary Charge Dudley Brierly Hill Hospitals INPP UK Unitary Charge Dudley Ridge Hill and Stour- Hospitals INPP UK Unitary Charge bridge Durham and Cleveland Police Gov Services HICL UK Unitary Charge Tactical Training Centre Durham Courts Gov Services INPP Canada Unitary Charge Durham PFI SPV 1 education INPP UK Unitary Charge Dutch High Speed Rail Link Roads HICL Netherlands Unitary Charge E18 Motorway Roads BBGI Norway Unitary Charge E18 Road Roads JLIF Finland Unitary Charge Ealing Care Homes Hospitals HICL UK Unitary Charge Ealing Schools, UK education HICL UK Unitary Charge East Down Colleges education BBGI UK Unitary Charge Ecole Centrale Supelec PPP education HICL France Unitary Charge Project, France Edinburgh Schools education JLIF, HICL UK Unitary Charge ELLAS Hospitals INPP UK Unitary Charge ELLAS 2 Hospitals INPP UK Unitary Charge ELLAS 3 Hospitals INPP UK Unitary Charge ELLAS 4 Hospitals INPP UK Unitary Charge Enfield Schools education JLIF UK Unitary Charge Enfield Street Lighting Gov Services JLIF UK Unitary Charge Essex PFI SPV 1 education INPP UK Unitary Charge Essex PFI SPV 2 education INPP UK Unitary Charge Exeter Crown Court, UK Gov Services HICL UK Unitary Charge Falkirk NPD Schools education HICL UK Unitary Charge Fife Schools 2 PPP education HICL UK Unitary Charge Fife Schools, UK education HICL UK Unitary Charge Forth Valley Royal Hospital Hospitals JLIF UK Unitary Charge Glasgow Hospital Hospitals HICL UK Unitary Charge Glasgow Schools education JLIF UK Unitary Charge Gloucester Fire and Rescue, UK Gov Services HICL UK Unitary Charge Gloucester Royal Hospital Hospitals BBGI UK Unitary Charge Golden Ears Bridge Roads BBGI Canada Unitary Charge Goscote Hospitals INPP UK Unitary Charge Greater Manchester Police Gov Services HICL, JlIF UK Unitary Charge Authority Groningen Tax Office Gov Services JLIF Netherlands Unitary Charge Harrow NRC Hospitals INPP UK Unitary Charge Haverstock School, UK education HICL UK Unitary Charge Health and Safety Executive Gov Services HICL UK Unitary Charge (HSE) Merseyside Headquarters Health and Safety Laboratory, education HICL UK Unitary Charge UK Helicopter Training Facility, UK education HICL UK Unitary Charge Hereford and Worcester Gov Services INPP UK Unitary Charge Highland School, Enfield education JLIF, HICL UK Unitary Charge Home Office Headquarters, UK Gov Services HICL UK Unitary Charge Irish Grouped Schools education HICL Ireland Unitary Charge Islington I Housing Gov Services JLIF, INPP UK Unitary Charge Islington II Housing Gov Services JLIF, INPP UK Unitary Charge Kelowna and Vernon Hospitals Hospitals BBGI, JLIF Canada Unitary Charge Kent PFI SPV 1 education INPP, BBGI, HICL UK Unitary Charge Kicking Horse Canyon Roads BBGI, HICL Canada Unitary Charge Kingston Hospital Hospitals JLIF UK Unitary Charge Kirklees Social Housing Gov Services JLIF UK Unitary Charge Kromhout Barracks Gov Services JLIF Netherlands Unitary Charge Lambeth Street Lighting Gov Services JLIF UK Unitary Charge Lancashire PFI SPV 1 education INPP UK Unitary Charge Lancashire PFI SPV 2 education INPP UK Unitary Charge Lancashire PFI SPV 2A education INPP UK Unitary Charge Lancashire PFI SPV 3 education INPP UK Unitary Charge Leeds Combined Secondary education JLIF UK Unitary Charge Schools Lewisham Hospital Hospitals HICL UK Unitary Charge Lewisham PFI SPV 1 education INPP UK Unitary Charge Lewisham PFI SPV 2 education INPP UK Unitary Charge Lewisham PFI SPV 3 education INPP UK Unitary Charge Lisburn College education BBGI UK Unitary Charge Liverpool and Sefton Clinics Hospitals BBGI UK Unitary Charge Liverpool Library Gov Services INPP UK Unitary Charge Long Bay Hospitals INPP Australia Unitary Charge

source: annual reports

A Publication of the EDHEC Infrastructure Institute-Singapore 57 Searching for a listed infrastructure asset class- June 2016

8. Appendices

Table 17: List of project finance SPVs in the listed PFI portfolio (continued)

Project name Sector Investor Country Revenue Source LUL Connect Roads JLIF UK Unitary Charge Luton PFI SPV 1 education INPP UK Unitary Charge M40 Motorway Roads JLIF UK Unitary Charge M6/M74 Project Roads JLIF UK Unitary Charge M80 DBFO Roads HICL,BBGI UK Unitary Charge Maesteg education INPP UK Unitary Charge Manchester School education HICL UK Unitary Charge Manchester Street Lighting Gov Services JLIF UK Unitary Charge Medway LIFT Hospitals HICL UK Unitary Charge Medway Police Gov Services HICL UK Unitary Charge Mersey Care Mental Health Hospitals BBGI UK Unitary Charge Hospital Metropolitan Police Specialist Gov Services HICL, JLIF UK Unitary Charge Training Centre Miles Platting Social Housing, Gov Services HICL, JLIF UK Unitary Charge UK Ministry of Defence Main Gov Services JLIF UK Unitary Charge Building Moray Schools education INPP UK Unitary Charge N17/18 Road, Ireland Roads HICL Ireland Unitary Charge Newcastle Hospital Hospitals JLIF UK Unitary Charge Newcastle Libraries education HICL UK Unitary Charge Newham Hospital Hospitals JLIF UK Unitary Charge Newham Schools education JLIF, INPP UK Unitary Charge Newport Schools education HICL UK Unitary Charge Newton Abbot Hospital Hospitals HICL UK Unitary Charge Norfolk Gov Services INPP UK Unitary Charge North East Fire and Rescue Gov Services JLIF UK Unitary Charge North Staffordshire Hospital Hospitals JLIF UK Unitary Charge North Swindon Schools education JLIF UK Unitary Charge North Tyneside Schools, UK education HICL UK Unitary Charge North Wales Police Authority Gov Services INPP UK Unitary Charge Northampton Mental Health Hospitals JLIF UK Unitary Charge Northampton Schools education INPP UK Unitary Charge Northeast Stoney Trail Roads BBGI Canada Unitary Charge Northwest Anthony Henday Roads HICL, BBGI Canada Unitary Charge Ring Road P3 Northwood MoD HQ Gov Services HICL UK Unitary Charge Norwich Area Schools PFI education HICL UK Unitary Charge Project Nottingham PFI SPV 1 education INPP UK Unitary Charge Nottingham PFI SPV 2 education INPP UK Unitary Charge NSW Schools education INPP Australia Unitary Charge Nuffield Hospital Hospitals HICL UK Unitary Charge Oldham Library education HICL UK Unitary Charge Oldham Secondary Schools PFI education HICL UK Unitary Charge Project Orange Hospital Hospitals INPP Australia Unitary Charge Oxford Churchill Oncology Hospitals HICL UK Unitary Charge Oxford Dunnock Way and East Hospitals INPP UK Unitary Charge Oxford Oxford John Radcliffe PFI Hospitals HICL UK Unitary Charge Hospital Pembury Hospital Hospitals JLIF UK Unitary Charge Perth and Kinross Schools education HICL UK Unitary Charge Peterborough Hospital Hospitals JLIF UK Unitary Charge Peterborough Schools education JLIF UK Unitary Charge Pforzheim Schools education INPP UK Unitary Charge Pinderfields and Pontefract Hospitals HICL UK Unitary Charge Hospitals, UK QFTO - Barrow Renewables INPP UK unknown QFTO - Gunfleet Sands Renewables INPP UK unknown QFTO - Lincs Renewables INPP UK unknown QFTO - Ormonde Renewables INPP UK unknown QFTO -Robin Rigg Renewables INPP UK unknown Queen Alexandra Hospital, Hospitals HICL UK Unitary Charge Portsmouth, UK Queen Elizabeth Hospital Hospitals JLIF UK Unitary Charge Queen’s (Romford) PFI Hospital Hospitals HICL UK Unitary Charge RD901 Road, France Roads HICL France Unitary Charge Realise Health (LIFT) Colchester Hospitals JLIF UK Unitary Charge Redbridge and Waltham Forest Hospitals HICL UK Unitary Charge LIFT Redcar and Cleveland Street Gov Services JLIF UK Unitary Charge Lighting Reliance Rail Rail Link INPP Australia unknown Renfrewshire Schools, UK education HICL UK Unitary Charge Rhonnda Cynon Taf Schools education HICL UK Unitary Charge Roseberry Park Hospital Hospitals JLIF UK Unitary Charge Royal Children’s Hospital Hospitals INPP Australia Unitary Charge Royal School of Military Gov Services HICL UK Unitary Charge Engineering PPP Project, UK Royal Women’s Hospital Hospitals BBGI Australia Unitary Charge

source: annual reports

58 A Publication of the EDHEC Infrastructure Institute-Singapore Searching for a listed infrastructure asset class - June 2016

8. Appendices

Table 18: List of project finance SPVs in the listed PFI portfolio (end)

Project name Sector Investor Country Revenue Source Royal Women’s Hospital Hospitals BBGI Australia Unitary Charge Salford and Wigan BSF Phase 1 education HICL, INPP UK Unitary Charge Salford and Wigan BSF Phase 2 education HICL, INPP UK Unitary Charge Salford Hospital, UK Hospitals HICL UK Unitary Charge Scottish Borders Schools education BBGI UK Unitary Charge Sheffield BSF, UK education HICL UK Unitary Charge Sheffield Hospital, UK Hospitals HICL UK Unitary Charge Sheffield Schools education HICL UK Unitary Charge Showgrounds Gov Services INPP Australia Unitary Charge Sirhowy Way Roads JLIF UK Unitary Charge Somerset PFI SPV 1 education INPP UK Unitary Charge South Ayrshire Schools, UK education HICL UK Unitary Charge South Bristol Community Hospitals INPP UK Unitary Charge Hospital South East London Police Gov Services HICL, JLIF UK Unitary Charge stations South Lanarkshire Schools education JLIF UK Unitary Charge South West Hospital, Hospitals HICL UK Unitary Charge Enniskillen Southwark PFI SPV 1 education INPP UK Unitary Charge Southwark PFI SPV 2 education INPP UK Unitary Charge St Thomas More School education INPP UK Unitary Charge Staffordshire LIFT Hospitals HICL UK Unitary Charge STaG PFI SPV 1 education INPP UK Unitary Charge STaG PFI SPV 2 education INPP UK Unitary Charge Stoke Mandeville Hospital, UK Hospitals HICL UK Unitary Charge Stoke on Trent and Stafford- Gov Services BBGI UK Unitary Charge shire Fire and Rescue Service Strathclyde Gov Services INPP UK Unitary Charge Surrey Street Lighting Gov Services JLIF UK Unitary Charge Sussex Custodial Services, UK Gov Services HICL UK Unitary Charge Tameside General Hospital Hospitals HICL UK Unitary Charge Tameside PFI SPV 1 education INPP UK Unitary Charge Tameside PFI SPV 2 education INPP UK Unitary Charge Tor Bank School education BBGI UK Unitary Charge Tower Hamlets Schools education INPP UK Unitary Charge Tyne and Wear Fire Stations Gov Services HICL UK Unitary Charge University of Bourgogne, education HICL UK Unitary Charge France University of Sheffield Project, Education HICL UK Unitary Charge UK Unna Administrative Centre Gov Services BBGI Germany Unitary Charge Vancouver General Hospital Hospitals JLIF Canada Unitary Charge Victoria Prisons Gov Services BBGI Australia Unitary Charge Wakefield Street Lighting Gov Services JLIF UK Unitary Charge Walsall Street Lighting Gov Services JLIF UK Unitary Charge Waltham Forest PFI SPV 1 education INPP UK Unitary Charge West Lothian Schools education HICL UK Unitary Charge West Middlesex Hospital, UK Hospitals HICL UK Unitary Charge Willesden Hospital Hospitals HICL UK Unitary Charge Wolvehampton PFI SPV 1 education INPP UK Unitary Charge Wolverhampton and Walsall Hospitals INPP UK Unitary Charge Wolverhampton PFI SPV 2 education INPP UK Unitary Charge Women’s College Hospital Hospitals BBGI Canada Unitary Charge Wooldale Centre for Learning, education HICL UK Unitary Charge UK Zaanstad Penitentiary Insti- Gov Services HICL Netherlands Unitary Charge tution, The Netherlands

source: annual reports

A Publication of the EDHEC Infrastructure Institute-Singapore 59 References

60 A Publication of the EDHEC Infrastructure Institute-Singapore Searching for a listed infrastructure asset class - June 2016

References

G Amenc, N., F. Goltz, A. Lodh, and L. Martellini (2014). Towards smart equity factor indices: Harvesting risk premia without taking unrewarded risks. Journal of Portfolio Management 40(4).

G Amenc, N., F. Goltz, L. Martellini, and P. Retkowsky (2010). Efficient indexation: An alternative to cap-weighted indices. EDHEC-Risk Institute Publications.

G Bender, J., R. Briand, F. Nielsen, and D. Stefek (2010, January). Portfolio of Risk Premia: A New Approach to Diversification. The Journal of Portfolio Management 36(2), 17–25.

G Bianchi, R. J., G. Bornholt, M. E. Drew, and M. F. Howard (2014, May). Long-term U.S. infrastructure returns and portfolio selection. Journal of Banking & Finance 42, 314– 325.

G Bird, R., H. Liem, and S. Thorp (2013, April). The Tortoise and the Hare: Risk Premium versus Alternative Asset Portfolios. Journal of Portfolio Management 39(3), 112–122.

G Bird, R., H. Liem, and S. Thorp (2014, May). Infrastructure: Real Assets and Real Returns. European Financial Management 20(4), 802–824.

G Bitsch, F. (2012). Do Investors Value Cash Flow Stability of Listed Infrastructure Funds?

G Blanc-Brude, F. (2013). Towards efficient benchmarks for infrastructure equity investments. EDHEC-Risk Institute Publications, 88.

G Blanc-Brude, F., M. Hasan, and T. Whittaker (2016, March). Revenue and dividend payout in privately held infrastructure investments. EDHEC Infrastructure Institute Publications March.

G Chen, H.-C., S.-L. Chung, and K.-Y. Ho (2011, May). The diversification effects of volatility-related assets. Journal of Banking & Finance 35(5), 1179–1189.

G Dechant, T. and K. Finkenzeller (2013). How much into infrastructure? Evidence from dynamic asset allocation. Journal of Property Research 30(2), 103–127.

G Eun, C. S., W. Huang, and S. Lai (2008). International Diversification with Large- and Small-Cap Stocks. Journal of Financial and Quantitative Analysis 43(02), 489–523.

G Fabozzi, F. J. and H. Markowitz (2011). The Theory and Practice of Investment Management (2nd ed.). Hoboken, New Jersey: John Wiley & Sons.

G Faff, R. (2001, June). An Examination of the Fama and French Three-Factor Model Using Commercially Available Factors. Australian Journal of Management 26(1), 1–17.

A Publication of the EDHEC Infrastructure Institute-Singapore 61 Searching for a listed infrastructure asset class- June 2016

References

G Fama, E. F. and K. R. French (1992, June). The Cross-Section of Expected Stock Returns. The Journal of Finance 47(2), 427–465.

G Fama, E. F. and K. R. French (1993, February). Common risk factors in the returns on stocks and bonds. Journal of Financial Economics 33(1), 3–56.

G Finkenzeller, K., T. Dechant, and W. Schäfers (2010). Infrastructure: a new dimension of real estate? An asset allocation analysis. Journal of Property Investment & Finance 28(4), 263–274.

G Gibbons, M. R., S. A. Ross, and J. Shanken (1989). A test of the efficiency of a given portfolio. Econometrica 57(5), 1121–52.

G Huberman, G. and S. Kandel (1987). Mean-Variance Spanning. The Journal of Finance 42(4), 873–888.

G Idzorek, T. and C. Armstrong (2009). Infrastructure and Strategic Asset Allocation: Is Infrastructure an Asset Class? Technical report, Ibbotson Associates.

G Ilmanen, A. and J. Kizer (2012, April). The Death of Diversification Has Been Greatly Exaggerated. The Journal of Portfolio Management 38(3), 15–27.

G Ince, O. S. and R. B. Porter (2006, December). Individual Equity Return Data From Thomson Datastream: Handle with Care. Journal of Financial Research 29(4), 463– 479.

G Jegadeesh, N. and S. Titman (1993, March). Returns to Buying Winners and Selling Losers: Implications for Stock Market Efficiency. The Journal of Finance 48(1), 65–91.

G Kan, R. and G. Zhou (2012). Tests of Mean-Variance Spanning. Annals of Economics and Finance 13, 139–187.

G Kroencke, T. A. and F. Schindler (2012, November). International diversification with securitized real estate and the veiling glare from currency risk. Journal of International Money and Finance 31(7), 1851–1866.

G MSCI (2014). MSCI World Index. http://www.msci.com/resources/factsheets/index_fact_sheet/msci- world-index.pdf.

G Newell, G., K. W. Chau, and S. K. Wong (2009). The significance and performance of infrastructure in China. Journal of Property Investment and Finance 27(2), 180–202.

G Newell, G. and H. W. Peng (2007). The significance and performance of retail property in Australia. Journal of Property Investment & Finance 25(2), 147–165.

62 A Publication of the EDHEC Infrastructure Institute-Singapore Searching for a listed infrastructure asset class - June 2016

References

G Newell, G. and H. W. Peng (2008). The role of US infrastructure in investment portfolios. Journal of Real Estate Portfolio Management 14(1), 21–34.

G Oyedele, J., A. Adair, and S. McGreal (2014). Performance of global listed infrastructure investment in a mixed asset portfolio. Journal of Property Research 31(1), 1–25.

G Peng, H. and G. Newell (2007). The significance of infrastructure in Australian investment portfolios. Pacific Rim Property Research Journal 13(4), 423–450.

G Petrella, G. (2005). Are Euro Area Small Cap Stocks an Asset Class? Evidence from Mean-Variance Spanning Tests. European Financial Management 11, 229–253.

G Rödel, M. and C. Rothballer (2012). Infrastructure as Hedge against Inflation—Factor Fantasy? The Journal of Alternative Investments 15(1), 110–123.

G Rothballer, C. and C. Kaserer (2012a). Is Infrastructure Really Low Risk? An Empirical Analysis of Listed Infrastructure Firms.

G Rothballer, C. and C. Kaserer (2012b). The Risk Profile of Infrastructure Investments: Challenging Conventional Wisdom. The Journal of Structured Finance 18(2), 95–109.

A Publication of the EDHEC Infrastructure Institute-Singapore 63 About the EDHEC Infrastructure Institute-Singapore

64 A Publication of the EDHEC Infrastructure Institute-Singapore Searching for a listed infrastructure asset class - June 2016

About the EDHEC Infrastructure Institute-Singapore

A Profound Knowledge Gap debt investment, with a focus on deliv- EDHECinfra addresses the profound knowledge gap Institutional investors have set their sights ering useful applied research in finance for faced by infrastructure on private investment in infrastructure investors in infrastructure. investors by collecting equity and debt as a potential avenue and standardising private towards better diversification, improved We aim to deliver the best available investment and cash flow liability-hedging and reduced drawdown data and running estimates of financial performance and state-of-the-art asset risk. risks of reference portfolios of privately- pricing and risk models to held infrastructure investments, and to create the performance Capturing these benefits, however, requires provide investors with important insights benchmarks that are answering a number of difficult questions: about their strategic asset allocation needed for asset allocation, prudential choices to infrastructure, as well as support regulation and the design 1. Risk-adjusted performance measures the adequate calibration of the relevant of new infrastructure are needed to inform strategic asset prudential frameworks. investment solutions. allocation decisions and monitoring performance; We are developing unparalleled access to 2. Duration and inflation hedging the financial data of infrastructure projects properties are required to understand and firms, especially private data that is the liability-friendliness of either unavailable to market participants infrastructure assets; or cumbersome and difficult to collect and 3. Extreme risk measures are in demand aggregate. from prudential regulators amongst others. We also bring advanced asset pricing and risk measurement technology designed Today none of these metrics is documented to answer investors’ information needs in a robust manner, if at all, for investors about long-term investment in privately- in privately-held infrastructure equity or held infrastructure, from asset allocation debt. This has left investors frustrated by to prudential regulation and performance an apparent lack of adequate investment attribution and monitoring. solutions in infrastructure. At the same time, policy-makers have begun calling for What We Do a widespread effort to channel long-term The EDHECinfra team is focused on three key savings into capital projects that could tasks: support long-term growth. 1. Data collection and analysis: we To fill this knowledge gap, EDHEC has collect, clean and analyse the private launched a new research platform, infrastructure investment data of the EDHECinfra, to collect, standardise and project’s data contributors as well as produce investment performance data for from other sources, and input it into infrastructure equity and debt investors. EDHECinfra’s unique database of infrastructure equity and debt investments Mission Statement and cash flows. We also develop data Our objective is the creation a global repos- collection and reporting standards that itory of financial knowledge and investment can be used to make data collection benchmarks about infrastructure equity and

A Publication of the EDHEC Infrastructure Institute-Singapore 65 Searching for a listed infrastructure asset class- June 2016

About the EDHEC Infrastructure Institute-Singapore

more efficient and reporting more Singapore Infrastructure Investment transparent. Institute (EDHEC infra) with a five-year This database already covers 15 years of research development grant. data and hundreds of investments and, as such, is already the largest dedicated database of infrastructure investment Sponsored Research Chairs information available. Since 2012, private sector sponsors have been supporting research on infrastructure 2. Cash flow and discount rate models: investment at EDHEC with several research Using this extensive and growing Chairs that are now under the EDHEC Infras- database, we implement and continue tructure Investment Institute: to develop the technology developed at EDHEC-Risk Institute to model the 1. The EDHEC/NATIXIS Research Chair on cash flow and discount rate dynamics of the Investment and Governance Charac- private infrastructure equity and debt teristics of Infrastructure Debt Instru- investments and derive a series of risk ments, 2012-2015 and performance measures that can 2. The EDHEC/Meridiam/Campbell Lutyens actually help answer the questions that Research Chair on Infrastructure Equity matter for investors. Investment Management and Bench- marking, 2013-2016 3. Building reference portfolios of 3. The EDHEC/NATIXIS Research Chair infrastructure investments: Using on Infrastructure Debt Benchmarking, the performance results from our 2015-2018 asset pricing and risk models, we can 4. The EDHEC/Long-Term Infrastructure report the portfolio-level performance Investor Association Research Chair on of groups of infrastructure equity or Infrastructure Equity Benchmarking, debt investments using categorisations 2016-2019 (e.g. greenfield vs brownfield) that are 5. The EDHEC/Global Infrastructure Hub most relevant for investors’ investment Survey of Infrastructure Investors’ decisions. Perceptions and Expectations, 2016

Partners of EDHECinfra Partner Organisations As well as our Research Chair Sponsors, Monetary Authority of Singapore numerous organisation have already recog- In October 2015, the Deputy Prime Minister nised the value of this project and have of Singapore, Tharman Shanmugaratnam, joined or are committed to join the data announced officially at the World Bank collection effort. They include: Infrastructure Summit that EDHEC would work in Singapore to create “usable bench- G The European Investment Bank; marks for infrastructure investors.” G The World Bank Group; G The European Bank for Reconstruction and Development; The Monetary Authority of Singapore G The members of the Long-Term Infras- is supporting the work of the EDHEC tructure Investor Association;

66 A Publication of the EDHEC Infrastructure Institute-Singapore Searching for a listed infrastructure asset class - June 2016

About the EDHEC Infrastructure Institute-Singapore

G Over 20 other North American, European This work has contributed to growing the and Australasian investors and infras- limited stock of investment knowledge in tructure managers. the infrastructure space.

EDHECinfra is also : Significant empirical findings already include: G A member of the Advisory Council of the World Bank’s Global Infrastructure G The first empirical estimates of Facility construction risk for equity and debt G An honorary member of the Long-term investors in infrastructure project Infrastructure Investor Association finance;

Origins and Recent Achievements G The only empirical tests of the statis- In 2012, EDHEC-Risk Institute created tical determinants of credit spreads in a thematic research program on infras- infrastructure debt since 2008, allowing tructure investment and established two controlling for the impact of market Research Chairs dedicated to long-term liquidity and isolating underlying risk investment in infrastructure equity and factors; debt, respectively, with the active support of the private sector. G The first empirical evidence of the diversification benefits of investing in Since then, infrastructure investment greenfield and brownfield assets, driven research at EDHEC has led to more than by the dynamic risk and correlation 20 academic publications and as many profile of infrastructure investments over trade press articles, a book on infrastructure their lifecycle; asset valuation, more than 30 industry and academic presentations, more than 200 G The first empirical documentation of the mentions in the press and the creation relationship between debt service cover of an executive course on infrastructure ratios, distance to default and expected investment and benchmarking. default frequencies;

G The first measures of the impact of Testament to the quality of its contributions embedded options in senior infras- to this debate, EDHEC infra’s research team tructure debt on expected recovery, has been regularly invited to contribute to extreme risk and duration measures; high-level fora on the subject, including G20 meetings. G The first empirically documented study of cash flow volatility and correlations Likewise, active contributions were made to in underlying infrastructure investment the regulatory debate, in particular directly using a large sample of collected data supporting the adaptation of the Solvency- covering the past fifteen years. 2 framework to long-term investments in infrastructure. Key methodological advances include:

A Publication of the EDHEC Infrastructure Institute-Singapore 67 Searching for a listed infrastructure asset class- June 2016

About the EDHEC Infrastructure Institute-Singapore

G A series of Bayesian approaches to modelling cash flows in long-term investment projects including predicting the trajectory of key cash flow ratios in a mean/variance plane;

G The first fully-fledged structural credit risk model of infrastructure project finance debt;

G A robust framework to extract the term structure of expected returns (discount rates) in private infrastructure investments using conditional volatility and initial investment values to filter implied required returns and their range at one point in time across heterogenous investors.

Recent contributions to the regulatory debate include:

G A parsimonious data collection template to develop a global database of infrastructure project cash flows;

G Empirical contributions to adapt prudential regulation for long-term investors.

68 A Publication of the EDHEC Infrastructure Institute-Singapore Infrastructure Research Publications at EDHEC

A Publication of the EDHEC Infrastructure Institute-Singapore 69 Searching for a listed infrastructure asset class- June 2016

Infrastructure Research Publications at EDHEC

EDHEC Publications

G Blanc-Brude, F., T. Whittaker and M. Hasan. Cash Flow Dynamics of Private Infras- tructure Debt (March 2016).

G Blanc-Brude, F., T. Whittaker and M. Hasan. Revenues and Dividend Payouts in Privately-Held Infrastructure Investments (March 2016).

G Blanc-Brude, F., and M. Hasan. The Valuation of Privately-Held Infrastructure Equity Investments (January 2015).

G Blanc-Brude, F., M. Hasan and O.R.H. Ismail. Performance and Valuation of Private Infrastructure Debt (July 2014).

G Blanc-Brude, F., Benchmarking Long-Term Investment in Infrastructure (June 2014).

G Blanc-Brude, F., and D. Makovsek. How Much Construction Risk do Sponsors take in Project Finance. (August 2014).

G Blanc-Brude, F. and O.R.H. Ismail. Who is afraid of construction risk? (March 2013)

G Blanc-Brude, F. Towards efficient benchmarks for infrastructure equity investments (January 2013).

G Blanc-Brude, F. Pension fund investment in social infrastructure (February 2012).

Books

G Blanc-Brude, F. and M. Hasan, Valuation and Financial Performance of Privately- Held Infrastructure Investments. London: PEI Media, Mar. 2015.

Peer-Reviewed Publications

G F. Blanc-Brude, S. Wilde, and T. Witthaker, “Looking for an infrastructure asset class Definition and mean-variance spanning of listed infrastructure equity proxies”, 2016 (forthcoming)

G Blanc-Brude, F., M. Hasan, and T. Witthaker, ”Benchmarking Infrastructure Project Finance - Objectives, Roadmap and Recent Progress”, Journal of Alternative Investments, 2016 (forthcoming)

G R. Bianchi, M. Drew, E. Roca and T. Whittaker, ”Risk factors in Australian bond returns”, Accounting & Finance, 2015

70 A Publication of the EDHEC Infrastructure Institute-Singapore Searching for a listed infrastructure asset class - June 2016

Infrastructure Research Publications at EDHEC

G Blanc-Brude, F. “Long-term investment in infrastructure and the demand for benchmarks,” JASSA The Finsia Journal of Applied Finance, vol. 3, pp. 57–65, 2014.

G Blanc-Brude, F. “Risk transfer, self-selection and ex post efficiency in public procurement: an example from UK primary and secondary school construction contracts,” Revue d’Economie Industrielle, vol. 141, no. 1st Quarter, pp. 149–180, 2013.

G Blanc-Brude, F. , H. Goldsmith, and T. Valila, “A comparison of construction contract prices for traditionally procured roads and public–private partnerships,” Review of Industrial Organization, vol. 35, no. 1-2, pp. 19–40, 2009, ISSN: 0889- 938X. DOI: 10.1007/s11151-009-9224-1.

G Blanc-Brude, F. , H. Goldsmith, and T. Valila, “Public-private partnerships in europe: an update,” EIB Economic & Financial Reports, p. 24, 2007.

G Blanc-Brude, F. and R. Strange, “How banks price loans to public-private partnerships: evidence from the european markets,” Journal of Applied Corporate Finance, vol. 19, no. 4, pp. 94–106, 2007.

G Blanc-Brude, F. , H. Goldsmith, and T. Valila, “Ex ante construction costs in the european road sector: a comparison of public-private partnerships and traditional public procurement,” EIB Economic & Financial Reports, vol. 2006/1, 2006.

G O. Jensen and F. Blanc-Brude, “The handshake: why do governments and firms sign private sector participation deals? evidence from the water and sanitation sector in developing countries,” World Bank Working Papers, Wold Bank Working Paper Series, no. October 2005, p. 25, 2006.

A Publication of the EDHEC Infrastructure Institute-Singapore 71 For more information, please contact: Karen Sequeira on +65 6438 0030 or e-mail: [email protected]

EDHEC Infrastructure Institute-Singapore EDHEC Business School Asia-Pacific One George Street - #07-02 Singapore 049145 Tel.: +65 6438 0030

edhec.infrastructure.institute